WO2021170015A1 - Method for acquiring point cloud data, and related device - Google Patents

Abstract

Disclosed are a method for acquiring point cloud data, and a related device, belonging to the technical field of artificial intelligence. The method comprises: determining a first scanning road section according to a first location point and a road topology model; determining three-dimensional model data of each of a plurality of virtual objects distributed in the first scanning road section; and determining point cloud data acquired by a laser detection apparatus at the first location point according to a laser scanning range, simulated on a vehicle, of the laser detection apparatus and the three-dimensional model data of each of the plurality of virtual objects. In the present application, a road topology model for a virtual space can be constructed in advance, so that a current road section to be scanned can be directly determined on the basis of the road topology model when subsequently determining point cloud data of a vehicle at a first location point, and there is no need to traverse all nodes in a K-D tree as is necessary in the related art, thereby reducing the data processing capacity during the process of acquiring point cloud data, and correspondingly improving the efficiency of acquiring the point cloud data.

Description

Method for obtaining point cloud data and related equipment

This application claims the priority of a Chinese patent application filed on February 25, 2020 with the application number 202010116696.X and the invention title "Method and Related Equipment for Obtaining Point Cloud Data", the entire content of which is incorporated into this application by reference middle.

Technical field

This application relates to the field of artificial intelligence technology, and in particular to a method and related equipment for obtaining point cloud data.

Background technique

With the development of artificial intelligence technology, self-driving vehicles have attracted more and more attention because of their ability to realize unmanned driving. Lidar is deployed on autonomous vehicles. Lidar can emit a large number of laser beams to the surrounding environment. Each laser beam is reflected by objects in the surrounding environment such as buildings. Based on the reflected laser beams, the surroundings of the autonomous vehicle can be simulated. Environment, so as to facilitate the automatic driving vehicle to plan the driving route according to the simulated surrounding environment. Before using an autonomous vehicle, it is necessary to perform a simulation test on the autonomous vehicle. The simulation test process is: constructing a virtual space, simulating the point cloud data of the position points projected by each laser beam onto the virtual object in the virtual space, the point cloud data includes the three-dimensional position information of the position points projected by the laser beam, According to the simulated point cloud data, the response of the autonomous vehicle is detected, so as to realize the performance test of the autonomous vehicle.

In related technologies, for a certain virtual space, a K-dimensional (K-D) tree is constructed according to the spatial position relationship of each virtual object in the virtual space. Each node in the K-D tree except the leaf node corresponds to a bounding box, which is used to indicate a part of the virtual space. Among the nodes other than the leaf nodes, the bounding box corresponding to each child node is a subset of the bounding box corresponding to the parent node to which the child node belongs. Each leaf node in the K-D tree corresponds to a virtual object in the virtual space, and the virtual object is a virtual object in the bounding box corresponding to the parent node to which the leaf node belongs. When acquiring simulated point cloud data, for each laser beam emitted by the lidar, each node is traversed in the order from the root node to the leaf node in the KD tree. For any node except the leaf node, if the laser beam is If the beam does not intersect the bounding box corresponding to the node, skip the node. If the laser beam intersects the bounding box corresponding to the node, continue to determine whether the laser beam intersects the bounding box corresponding to the child node of the node until it traverses to the leaf node, thereby determining the virtual object that intersects the laser beam, Then obtain the point cloud data of the laser beam on the virtual object. All the laser beams are processed as described above, and the point cloud data of each laser beam can be obtained.

In the above process of obtaining point cloud data, all laser beams need to traverse the K-D tree, resulting in the process of obtaining point cloud data that consumes a lot of computing resources.

Summary of the invention

This application provides a method and related equipment for acquiring point cloud data, which can improve the efficiency of acquiring point cloud data. The technical solution is as follows:

In the first aspect, a method for obtaining point cloud data is provided, and the method is applied to a computer device. In this method, when the vehicle is at the first position point in the virtual space, determine the first scanned road section according to the first position point and the road topology model, where the road topology model is used to indicate the road topology in the virtual space; The three-dimensional model data of each virtual object among the multiple virtual objects distributed in the first scanning section; according to the laser scanning range of the laser detection device simulated on the vehicle and the three-dimensional model data of each virtual object in the multiple virtual objects , Determine the point cloud data obtained by the laser detection device at the first position point.

In this application, it is considered that the autonomous driving vehicle only needs to determine the objects near the driving road during the driving process, therefore, in order to reduce the amount of data processing in the process of determining the point cloud data. For the virtual space to be simulated, a road topology model for the virtual space can be constructed in advance. The road topology model is used to indicate the road topology in the virtual space. In this way, when subsequently determining the point cloud data when the vehicle is at the first position point, the current road segment to be scanned can be directly determined based on the road topology model, without the need to traverse all nodes in the KD tree as in related technologies, thereby reducing The amount of data processing in the process of obtaining point cloud data, correspondingly, improves the efficiency of obtaining point cloud data.

According to the first aspect, in a possible implementation of the present application, the road topology model includes multiple road nodes and road information of each road node in the multiple road nodes. In this scenario, the implementation of determining the first scanned road segment according to the first location point and the road topology model may be: determining the road node corresponding to the first location point from the multiple road nodes according to the road information of each road node; The first scanning road segment is determined according to the road node corresponding to the first position point.

In order to quickly determine the first scanned road segment, the road segments in the virtual space are stored in the road topology model in the form of road nodes in advance, so that the first scanned road segment can be quickly determined based on the first location point, thereby improving the acquisition The efficiency of point cloud data.

According to the first aspect, in a possible implementation of the present application, the road topology model includes multiple road nodes and virtual object information of each road node in the multiple road nodes. Wherein, the virtual object information of the first road node among the multiple road nodes includes three-dimensional model data of each virtual object distributed on the first road node, and the first road node is any one of the multiple road nodes. In this scenario, the way to determine the three-dimensional model data of each virtual object among the multiple virtual objects distributed on the first scanning road segment may be: obtaining the virtual object information distributed on the first scanning road segment from the virtual object information of each road node The three-dimensional model data of each virtual object in the multiple virtual objects.

In order to further improve the efficiency of obtaining point cloud data, the relevant information of the virtual objects distributed on each road section can also be stored in the road topology model in advance. In this way, after the first scanning road section is determined, the three-dimensional model data of each virtual object among the multiple virtual objects distributed in the first scanning road section can be directly determined according to the road topology model, so as to improve the efficiency of obtaining point cloud data.

According to the first aspect, in a possible implementation of the present application, the distance between the virtual object distributed on the first road node and the center line of the first road node is within the first distance threshold.

Considering that during the driving of an autonomous vehicle, virtual objects on both sides of the road that are particularly far away from the autonomous vehicle will not have any impact on the driving of the autonomous vehicle. Therefore, it is also possible to filter virtual objects that are particularly far apart on both sides of the road. In this way, in the road topology model, the virtual objects distributed at each road node only include the virtual objects that are within the first distance threshold from the center line of the corresponding road node. This simplifies the road topology model, which can reduce the amount of subsequent calculations for obtaining point cloud data based on the road topology model, and improve the efficiency of obtaining point cloud data.

According to the first aspect, in a possible implementation of the present application, the three-dimensional model data of the first virtual object includes the three-dimensional position information of each of the multiple surface points of the first virtual object, and the three-dimensional position information includes the surface The height of the point, the first virtual object is any one of the multiple virtual objects. In this scenario, the height of the three-dimensional model data of the multiple virtual objects distributed on the first scanning section is within the height threshold.

Considering that during the driving of an autonomous vehicle, virtual objects near the road that are particularly high from the ground will not have any impact on the driving of the autonomous vehicle. For example, the high-rise buildings on both sides of the road will not have any impact on the driving of autonomous vehicles. Therefore, it is also possible to filter virtual objects near the road or parts on the virtual objects that are particularly high from the ground. At this time, in the road topology model, there is no three-dimensional position information of the surface points whose height exceeds the height threshold in the three-dimensional model data of each of the multiple virtual objects distributed in the road nodes. In this way, the road topology model can also be simplified, which can reduce the amount of subsequent calculations for obtaining point cloud data according to the road topology model, thereby improving the efficiency of obtaining point cloud data.

According to the first aspect, in a possible implementation of the present application, the three-dimensional model data of the first virtual object among the multiple virtual objects includes the three-dimensional position information of the bounding box of the first virtual object, and the bounding box of the first virtual object Refers to the geometric body surrounding the first virtual object, and the first virtual object is any one of the multiple virtual objects. In this scenario, according to the laser scanning range of the laser detection device simulated on the vehicle and the three-dimensional model data of each virtual object in the multiple virtual objects, the realization of the point cloud data obtained by the laser detection device at the first position point is determined The method may be: according to the three-dimensional position information of the bounding box of the first virtual object, the orientation of the vehicle, and the first position point, determine from the laser scanning range the angular range of the laser beam emitted by the laser detection device covering the first virtual object; according to The three-dimensional model data of the first virtual object determines the intersection point of the first laser beam on the first virtual object, and uses the three-dimensional position information of the intersection point as the point cloud data corresponding to the first laser beam. The first laser beam is any angle range. A laser beam.

In this application, acquiring the point cloud data of the laser detection device at the first position point is essentially to determine the intersection of each laser beam on the virtual object. Therefore, the point cloud data of the laser detection device at the first position point can be determined through the foregoing implementation manner.

According to the first aspect, in a possible implementation manner of the present application, the implementation manner of determining the intersection point of the first laser beam on the first virtual object according to the three-dimensional model data of the first virtual object may be: When the object is a short-distance virtual object, according to the three-dimensional position information of each surface point in the three-dimensional model data of the first virtual object, a plurality of first facets of the first virtual object are determined, and each first facet refers to the first facet. A surface formed by three or more surface points among the various surface points of a virtual object. A short-distance virtual object refers to a virtual object whose distance from the first position point is within the second distance threshold; The first surface element that intersects the first laser beam among the first surface elements uses the intersection point between the first laser beam and the intersecting first surface element as the intersection point of the first laser beam on the first virtual object.

According to the first aspect, in a possible implementation manner of the present application, the implementation manner of determining the intersection point of the first laser beam on the first virtual object according to the three-dimensional model data of the first virtual object may be: When the object is a long-distance virtual object, according to the three-dimensional position information of the bounding box of the first virtual object, multiple second bins of the bounding box are determined. The long-distance virtual object refers to the distance between the first position and the The virtual object between the second distance threshold and the first distance threshold, the first distance threshold is greater than the second distance threshold; the second surface element that intersects the first laser beam among the plurality of second surface elements is determined, and the first laser beam The intersection point with the second intersecting surface element serves as the intersection point of the first laser beam on the first virtual object.

Taking into account that during the driving of an autonomous vehicle, it is necessary to clarify the detailed three-dimensional structure of nearby objects to avoid traffic accidents. For distant objects, you only need to know the approximate outline to guide the current planning. Therefore, for the first laser beam in the angular range, in the process of determining the intersection point of the first laser beam on the first virtual object according to the three-dimensional model data of the first virtual object, the distance from the first position point may be different. The virtual object uses different processing methods to determine the intersection point.

According to the first aspect, in a possible implementation of the present application, before the intersection point of the first laser beam on the first virtual object is determined according to the three-dimensional model data of the first virtual object, if the first laser beam is projected If there is no other virtual object blocking the first virtual object in the direction, the operation of determining the intersection point of the first laser beam on the first virtual object is performed.

Taking into account the occlusion between virtual objects, the same laser beam may theoretically be projected on different virtual objects, but in fact, each laser beam can only collect one point cloud data. Therefore, when determining the point cloud data, the occlusion between virtual objects can also be considered, which can improve the consistency between the point cloud data acquired in the virtual space and the point cloud data acquired in the real world.

According to the first aspect, in a possible implementation of the present application, in the method, if there are other virtual objects that block the first virtual object in the projection direction of the first laser beam, and the other virtual objects are non-strict For the obstructed object, the operation of determining the intersection point of the first laser beam on the first virtual object is performed, and the non-strictly obstructed object refers to an object with a through hole through which the laser beam can penetrate. Correspondingly, after determining the intersection point of the first laser beam on the first virtual object, if the first laser beam has no intersection point on other virtual objects, the three-dimensional intersection point of the first laser beam on the first virtual object The position information is used as the point cloud data corresponding to the first laser beam.

In addition, in the process of determining whether there are other virtual objects that block the first virtual object in the projection direction of the first laser beam, the other virtual objects can be treated as a simple model of impermeable laser beam. However, in the real world, for objects with complex models such as trees, there are through holes in the trees that can pass through the laser beam. Therefore, even if the bounding box corresponding to the tree blocks the first laser beam, the laser beam may still be Will be projected onto the first virtual infinite body. At this time, the point cloud data can be determined through the foregoing implementation method, so as to improve the consistency between the point cloud data obtained in the virtual space and the point cloud data obtained in the real world.

According to the first aspect, in a possible implementation manner of the present application, according to the virtual object information of each road node, the implementation manner of determining the three-dimensional model data of each virtual object among the multiple virtual objects distributed in the first scanning road section It can be: divide the first scanning road section into multiple fan-shaped areas centered on the first position point, and obtain the boundary information of each fan-shaped area; arrange the multiple fan-shaped areas in order according to the scanning direction of the laser detection device; The first fan-shaped area, according to the boundary information of the first fan-shaped area, obtain the three-dimensional model data of the virtual object located in the first fan-shaped area from the virtual object position information of each road node; for the i-th fan-shaped area after arrangement , Based on the scanning speed of the laser detection device and the moving speed of the vehicle, determine the movement displacement of the vehicle during the scanning of the laser detection device from the first sector area to the i-th sector area, i is greater than or equal to 2, and less than or equal to The positive integer of the number of divided sector areas; according to the movement displacement, update the boundary information of the i-th sector area; according to the updated boundary information of the i-th sector area, obtain the location information from the virtual object position information of each road node The three-dimensional model data of the virtual object in the first sector area.

As the lidar scans, the vehicle is moving forward. If the movement displacement of the vehicle is relatively large during the horizontal scanning of the lidar, in this case, in order to improve the consistency between the point cloud data obtained by the lidar during the simulation test and the point cloud data obtained in the real world , The three-dimensional model data of each virtual object among the multiple virtual objects distributed on the first scanning road section and its two sides can be determined in a sub-regional manner.

According to the first aspect, in a possible implementation of the present application, the point cloud data acquired by the laser detection device at the first position point includes point cloud data corresponding to each laser beam emitted by the laser detection device. At this time, after determining the point cloud data obtained by the laser detection device at the first position point, the point cloud data corresponding to each laser beam is cached; when the vehicle is at the second position point in the virtual space, according to the second position Point and road topology model determine the second scan section; determine multiple virtual objects distributed on the second scan section; if there are multiple virtual objects distributed on the first scan section and multiple virtual objects distributed on the second scan section The same virtual object, and the relative distance between the same virtual object and the vehicle does not change between the first position point and the second position point, the point cloud data corresponding to the laser beam projected on the same virtual object in the buffer is used as the corresponding laser Point cloud data corresponding to the light beam at the second position point.

In the simulation test, as the vehicle moves in the virtual space, it is necessary to update the point cloud data at each location in time. In the process of vehicle movement, there may be some objects around the vehicle and the relative position between the vehicle has not changed. In this case, for these virtual objects, there is no need to determine the point cloud data again. Therefore, in this application, the amount of calculation required to determine the point cloud data can be reduced according to the relative state between the vehicle and the surrounding virtual objects.

In a second aspect, a device for acquiring point cloud data is provided, and the device has the function of realizing the method and behavior of acquiring point cloud data in the above-mentioned first aspect. The device includes at least one module, and the at least one module is used to implement the method for obtaining point cloud data provided in the above-mentioned first aspect.

In a third aspect, a computer device is provided. The computing device includes a memory and a processor, the memory is used to store computer instructions, and the processor is used to read the computer instructions to execute the acquisition described in the first aspect. Point cloud data method.

In a fourth aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores instructions that, when run on a computer, cause the computer to execute the method for obtaining point cloud data as described in the first aspect. .

In a fifth aspect, a computer program product containing instructions is provided, which when running on a computer, causes the computer to execute the method for obtaining point cloud data as described in the first aspect.

The technical effects obtained by the above second, third, fourth and fifth aspects are similar to the technical effects obtained by the corresponding technical means in the first aspect, and will not be repeated here.

Description of the drawings

Fig. 1 is a schematic diagram of a laser beam emitted by a lidar provided by an embodiment of the present application;

2 is a schematic diagram of another laser beam emitted by a lidar provided by an embodiment of the present application;

FIG. 3 is a schematic diagram of another laser beam emitted by a lidar provided by an embodiment of the present application;

FIG. 4 is a schematic diagram of a simulated laser beam intersecting a triangular surface element provided by an embodiment of the present application;

FIG. 5 is a schematic diagram of the architecture of a simulation system provided by an embodiment of the present application;

Fig. 6 is a flowchart of a method for obtaining point cloud data provided by an embodiment of the present application;

FIG. 7 is a schematic diagram of a road topology model provided by an embodiment of the present application;

FIG. 8 is a schematic diagram of a bounding box of a virtual object provided by an embodiment of the present application;

FIG. 9 is a schematic diagram of a high-rise building provided by an embodiment of the present application;

FIG. 10 is a schematic diagram of a road node provided by an embodiment of the present application;

FIG. 11 is a schematic diagram of dividing a sector area according to an embodiment of the present application;

FIG. 12 is a schematic diagram of determining an angle range provided by an embodiment of the present application;

FIG. 13 is a schematic diagram of a target vehicle in front of an autonomous driving vehicle provided by an embodiment of the present application;

FIG. 14 is a schematic diagram of projection of a laser beam provided by an embodiment of the present application;

FIG. 15 is a schematic diagram of a device for obtaining point cloud data provided by an embodiment of the present application;

FIG. 16 is a schematic structural diagram of a computer device provided by an embodiment of the present application.

Detailed ways

In order to make the purpose, technical solutions, and advantages of the present application clearer, the implementation manners of the present application will be described in further detail below in conjunction with the accompanying drawings.

Before explaining the embodiments of the present application in detail, the application scenarios of the embodiments of the present application are first explained.

As the society's demand for driving intelligence, economy, safety and other aspects increases, autonomous driving technology has become one of the key development directions of the automotive industry, and it has received more and more attention from OEMs and Internet companies. Among them, autonomous vehicles (in English may be autonomous vehicles or self-piloting automobiles), also known as unmanned vehicles, computer-driven vehicles, or wheeled mobile robots, are intelligent vehicles that realize unmanned driving through a computer system. Self-driving vehicles have a history of decades in the 20th century. At the beginning of the 21st century, they were approaching practicality.

In the architecture of autonomous vehicles, the sensing layer is compared to the “eyes” of autonomous vehicles. The sensing layer includes vision sensors such as vehicle cameras and radar sensors such as vehicle millimeter wave radar, vehicle laser radar, and vehicle ultrasonic radar. Among them, lidar has been considered as a necessary foundation for realizing autonomous driving. In addition, an autonomous vehicle equipped with a lidar or a simulated vehicle equipped with a simulated lidar model in a simulation can also be called a self-driving vehicle.

In order to facilitate the subsequent description, the working principle of the lidar is explained here.

Lidar is a system that integrates laser, GPS global positioning and inertial measurement devices. Lidar can obtain data and generate accurate digital models. The combination of these three technologies can accurately locate the position where the laser beam strikes. Compared with other detection products, lidar can give full play to the advantages of precision, speed and efficiency. The biggest feature of lidar is that it can generate three-dimensional position information, and even quickly determine the position, size, external shape and even material of an object, so that an accurate digital model can be generated.

Lidar is a radar system that detects characteristics such as the position and speed of a target by emitting a laser beam. Lidar emits a large number of laser beams to the surrounding environment in a short time, and calculates the distance to surrounding objects by measuring the flight time of the reflected laser beam. Lidar can construct a 3D map of the surrounding environment in an instant, and has the advantages of high measurement accuracy and good directionality.

Lidar continuously emits laser beams to the surrounding environment when it is working, and calculates the distance between the vehicle and the point where the laser beam is projected by measuring the transmission time of the emitted laser beam, and then quickly creates point cloud data of the surrounding environment. Simply put, Lidar is a device that quickly creates point cloud data of various points in the surrounding environment. The point cloud data includes the three-dimensional coordinates (XYZ) and the laser reflection intensity (Intensity) of the point where the laser beam is projected. The intensity of laser reflection is related to the surface material, roughness, incident angle direction of the corresponding position, the emission energy of the instrument, and the wavelength of the laser beam.

Lidar can create up to 1.5 million point cloud data per second. This ability to quickly create point cloud data makes Lidar useful in areas such as autonomous driving. For example, in autonomous driving, point cloud data is used as the output data of lidar, which can provide surrounding environment information to autonomous vehicles, and serve as the development and inspection of algorithms for perception, planning, decision-making, and information fusion in the process of autonomous driving. In addition, point cloud data can also be used as training data for these algorithms.

Fig. 1 is a schematic diagram of a laser beam emitted by a lidar provided by an embodiment of the present application. Each concentric circle in Figure 1 corresponds to the point cloud data obtained after a group of laser scanning in the horizontal direction. Each concentric circle can also be called a point cloud concentric circle. For two groups of adjacent lasers, the vertical separation angle is constant. Therefore, the farther the distance between the adjacent two groups of lasers is, the greater the distance between the concentric circles of the point cloud scanned by the adjacent lasers.

The position information indicated by the part of the point cloud data in Figure 1 forms a concentric circle shape. That is because there are no obstacles around the autonomous vehicle. The laser beam hits the ground to form a group of concentric circles. The specific principle is shown in Figure 2. . If there are obstacles around the autonomous vehicle, the corresponding shape of the point cloud data is the shape of the obstacle, as shown in Figure 1 for the point cloud data on the buildings surrounding the autonomous vehicle.

The following explains several parameters involved in the working process of lidar.

Azimuth angle: The scanning range of the laser beam when the lidar is working is called the azimuth angle (field of view, FOV) of the lidar. The azimuth angle includes the horizontal azimuth angle and the vertical azimuth angle. The horizontal azimuth angle refers to the scanning angle of the lidar in the horizontal direction. The vertical azimuth angle refers to the scanning angle of the lidar in the vertical direction.

At present, most lidar systems use rotating lenses. As shown in Figure 2, the main part of the lidar is fixed on the base of the rotating motor, and it rotates continuously during work to scan the surrounding 360°. In other words, the horizontal azimuth of these lidars is 360°.

The vertical azimuth angle refers to the scanning angle of the lidar in the vertical direction, generally within 40°. As shown in Figure 3, it is a schematic diagram of a certain laser radar laser scanning line, and its vertical azimuth angle is -16°～7°.

Data update frequency: Because the lidar is constantly scanning in a circle when it is working, and during the lidar scanning process, the self-driving vehicle is moving. Therefore, the point cloud data of the laser beam emitted by the laser radar in the same azimuth is updated every time the laser radar scans one circle. For example, the scanning frequency of the lidar is 10 Hz, that is, the lidar rotates 10 times per second. Therefore, the point cloud data of the laser beam emitted from the same direction will be updated 10 times per second.

Angular resolution: The angular resolution of lidar is divided into horizontal angular resolution and vertical angular resolution. The horizontal angular resolution refers to the minimum degree of separation between scan lines in the horizontal direction. It changes with the change of the scanning frequency, which is also the rotation speed of the lidar. The faster the rotation speed, the greater the spacing of the scanning lines in the horizontal direction and the greater the horizontal angular resolution. The vertical angular resolution refers to the degree of separation between two scan lines in the vertical direction. As shown in Fig. 3, the vertical angular resolution of the lidar is 1° between the 1-5 lines, 0.33° between the 6-30 lines, and 1° between the 31-40 lines.

Ranging range: The ranging range of lidar used in the field of autonomous driving is generally about 100-200m.

Data rate: The unit of the number of point cloud data generated by lidar per second is points/sec. Among them, the number of point cloud data generated per second may also be referred to as data rate (Data Rate) or point cloud measurement (Measurement Points). For example: a laser radar with a scanning frequency of 20 Hz, the horizontal angular resolution is 0.45° (800 scans per laser per circle). Therefore, the number of point cloud data generated per second is: 40×20×800=640,000 points/sec.

In addition, a large amount of point cloud data scanned by lidar is sent to other devices that need to use point cloud data in the format of user datagram protocol (UDP) data packets. Take Hesai Pandar40P (a type of lidar) as an example. The UDP data packet sent by Pandar40P is 1304 bytes in total, including a header of 42 bytes and a data block interval of 1262 bytes. All multi-byte values are in Little Endian and are unsigned integers. Little-endian byte order refers to a data storage method in which the low-order byte is first and the high-order byte is last, and will not be described in detail here.

Among them, the data block interval in a UDP data packet includes 10 data blocks, and each data block has a length of 124 bytes, which represents a complete set of point cloud data. The 124-byte space in the data block includes: 2 bytes of flag bits, 2 bytes of horizontal rotation angle information, 40 groups of laser beam information, each group of laser beam information contains 2 bytes of distance information and 1 byte of information Strength information. In addition, the data block interval also includes 22 bytes of additional information, which includes sensor temperature status information, motor speed information, time stamp information (time stamp information is used to indicate the current packing time of the data block, in microseconds), etc. data.

The above content is used to explain the working principle of lidar. At present, in order to ensure the safety of self-driving vehicles, it is possible to test the self-driving vehicles before using them.

At present, the industry has passed a large number of road tests to cover as many actual driving environments and traffic conditions as possible. However, according to statistical calculations, based solely on this method, the autonomous vehicle under test needs to pass billions of kilometers of test verification to achieve a certain degree of verification. Obviously, it is very difficult, time-consuming and laborious to complete a test on the order of a kilometer through a road test in the real world. At the same time, the real road test also has the problem of high risk and low occurrence rate.

Therefore, in countries where the automobile industry is developed, host companies, first-tier suppliers, and equipment suppliers often use simulation test methods to verify the functions of autonomous vehicles. Through the use of simulation software, based on the real traffic environment, a virtual space is generated or reproduced in the simulation software to test whether the autonomous vehicle can correctly recognize the surrounding environment in the virtual space, and make timely and accurate responses and take appropriate driving behavior. The industry has reached a consensus that only by combining simulation tests with actual road tests can comprehensive, systematic and effective verification results be obtained. Therefore, it is more and more important to conduct simulation tests on autonomous vehicles.

The simulation test of the autonomous vehicle is mainly through the simulation system, the point cloud data output by the detection device on the autonomous vehicle is simulated, and the point cloud data is transmitted to the autonomous vehicle to observe whether the response of the autonomous vehicle is correct. Because lidar is the basic detection device for autonomous driving, simulation testing of autonomous vehicles can refer to determining the point cloud data of lidar in virtual space.

In the real world, lidar uses laser beams to detect surrounding real objects to obtain spatial information of surrounding real objects. In the simulation test, all environments are virtual, including virtual vehicles, virtual traffic conditions, virtual roads, buildings and other static objects. Therefore, in the simulation test, the lidar is required to detect the virtual environment and output point cloud data.

In the virtual space, all objects are represented by three-dimensional models. The three-dimensional model is the input data of computer graphics. It not only describes the specific geometric shape of the object in the real world, but also records the surface information such as the material, color, and texture of the surface of the object.

The three-dimensional model describes the surface geometry of an object through geometric elements such as points, lines, and surfaces and the topological relationships between them. The surface properties of the object are described by adding information such as material, color, and texture to the various bins that make up the surface of the object. The current industry standard 3D rendering engine Open Graphics Library (open graphics library, openGL) uses a 3D model based on boundary representation to describe 3D objects.

The three-dimensional model expresses the surface of the object in the form of polygons. For the vertices of the polygon or the polygon itself, there will be corresponding colors, normal vectors, and textures to describe other attributes of the surface of the object besides the shape. Specifically, a three-dimensional model consists of two parts: geometric structure and surface properties. For the convenience of description, the three-dimensional model can be remembered as:

Model={GS,A}

Model is used to indicate the three-dimensional model, Gs is used to indicate the data set of the geometric structure of the three-dimensional model, and A is used to indicate the data set of the surface properties of the three-dimensional model. The embodiments of this application do not involve surface properties, so A is not described in detail here.

GS is represented by points, lines, polygons and the topological relationship between them. Therefore, the above Gs can be expressed as:

GS={V,Tri}

V is used to indicate the collection of all points in the three-dimensional model. V can be expressed as follows:

V={Vi(xi,yi,zi)|i=1,...,N}

Among them, N is the number of points in the three-dimensional model. In GS, the surface of a three-dimensional model is represented by a polygon set with three-dimensional points in V as vertices. That is, for the surface in the three-dimensional model, it is decomposed into several polygons. Use these polygons to express the surface of the model (the surface can also be called a facet). Since in the three-dimensional model, the vertices of the polygon all come from the vertex set V, a polygon with k vertices is represented by the k vertices in V arranged in a counterclockwise manner. Since the coordinates of the vertices are already included in the set V, when representing a polygon, only the index of the vertices in V can be specified. Theoretically, the number of vertices of the polygon used to express the surface can be arbitrary. But in practice, when the number of vertices is more than 3, concave polygons will appear, and openGL's rendering of concave polygons is unstable. In addition, due to the error of the vertex coordinates, when the number of vertices of the polygon is more than 3, it is possible that the vertices of the polygon are not in the same plane, which will cause the polygon to not be displayed. Therefore, in actual use, although the rendering engine provides support for arbitrary polygons, considering stability and efficiency, only a set of triangles can be used to express the surface of a three-dimensional model. For example, in a three-dimensional model, the set of all triangles is Tri, which can be expressed as follows:

Tri={tri(ai,bi,ci)|V _ai ,V _bi ,V _ci ∈V,i=1,...,M}

ai bi ci is the index in the vertex set V of the three vertices of the i-th triangle arranged in counterclockwise order in Tri. Together, V and Tri fully express the geometric structure of the 3D model surface.

In the simulation test, the calculation of the point cloud data is obtained by calculating the intersection of the simulated laser beam and the triangular facet in the three-dimensional model. FIG. 4 is a schematic diagram of a simulated laser beam intersecting with a triangular surface element provided in an embodiment of the present application. As shown in Fig. 4, the arrowed line is the simulated laser beam, and V1 to V8 are the 8 vertices in the three-dimensional model of the simulated object. The laser beam has an intersection with the facets V ₁ V ₂ V ₅ and V ₂ V ₃ V _{6 of the} three-dimensional model. Since the distance between the intersection point on the face element V ₁ V ₂ V ₅ and the origin of the laser beam is less than the distance between the intersection point on the face element V ₂ V ₃ V ₆ and the origin, the intersection point on the _{face element V 1} V ₂ V ₅ is the location of the ray. The irradiated position point of the simulated laser beam on the object corresponding to the three-dimensional model, and the three-dimensional position information of the intersection point is the point cloud data scanned when the laser beam simulated by the ray is projected on the object. Therefore, simulating the laser radar output point cloud data in the simulation space is to determine the intersection information of all laser beams and the surrounding virtual virtual bodies.

The embodiment of the present application provides a method for obtaining point cloud data, which can obtain the point cloud data of the lidar in a virtual space. The point cloud data can be used for the development and inspection of algorithms such as verification, perception, planning, decision-making, and information fusion of autonomous driving systems.

FIG. 5 is a schematic diagram of the architecture of a simulation system provided by an embodiment of the present application. As shown in FIG. 5, the simulation system 500 includes a simulation platform 501 and an autonomous driving platform 502. The simulation platform 501 and the autonomous driving platform 502 may be connected in a wired or wireless manner for communication.

The simulation platform 501 may run on a computer, and is used to obtain point cloud data through the method provided in the embodiments of the present application, and send the point cloud data to the automatic driving platform 502 through a network. The autonomous driving platform 502 controls the autonomous vehicle to respond according to the point cloud data to test the performance of the autonomous vehicle. Among them, the self-driving platform can run on self-driving vehicles.

In addition, when there is no communication between the simulation platform 501 and the autonomous driving platform 502, after the simulation platform 501 obtains the point cloud data, it can also transmit the point cloud data to the autonomous driving platform 502 through a storage medium. No more detailed description.

In addition, as shown in FIG. 5, the simulation system 500 may also include an algorithm platform 503, which can perform other algorithm processing based on the point cloud data obtained by the simulation platform 501, which is also not described in detail here.

Next, the method for obtaining point cloud data provided by the embodiments of the present application will be explained in detail. FIG. 6 is a flowchart of a method for obtaining point cloud data provided by an embodiment of the present application, and the method may be applied to the simulation platform in FIG. 5. As shown in Figure 6, the method includes the following steps:

Step 601: In the case that the vehicle is at the first position point in the virtual space, determine the first scanning road segment according to the first position point and the road topology model, where the road topology model is used to indicate the road topology in the virtual space.

In the embodiments of the present application, considering that the autonomous vehicle only needs to determine objects near the driving road during the driving process, in order to reduce the amount of data processing in the process of determining point cloud data, the virtual space to be simulated can be A road topology model for the virtual space is constructed in advance. The road topology model is used to indicate the road topology in the virtual space. In this way, when determining the point cloud data when the vehicle is at the first location point later, step 601 can be used to determine the current road segment to be scanned based on the road topology model, without traversing all the spaces around the first location point.

The above-mentioned first position point may be the current position of the laser detection device simulated on the vehicle. The laser detection device may be a laser radar, or other types of laser detection devices, which will not be illustrated here.

In a possible implementation manner, the road topology model may include multiple road nodes (road) and road information of each road node. Each road node corresponds to a section of road, that is, in the embodiment of the present application, each road node is essentially used to indicate a section of road. Among them, the road information of each road node is used to indicate various attributes of the corresponding road. The use of the road information of each road node to indicate various attributes of the corresponding road means that the specific location of the corresponding road in the virtual space can be clarified through the road information corresponding to the road node. That is, the topological relationship between each road in the virtual space can be constructed through the road information of each road node in the road topology model. For example, the road information of each road node may include the identification of the road node, the length of the road corresponding to the road node, and the boundary position information of the road corresponding to the road node (the boundary position information may include the starting point position information of the corresponding road and End position information, the connection relationship between the road corresponding to the road node and other roads, etc.).

Since the road information of each road node can indicate the specific location of the corresponding road, based on the above road topology model, the implementation of determining the first scanning road segment according to the first location point and the road topology model in step 601 can be: according to each For the road information of the road node, the road node corresponding to the first location point is determined from the multiple road nodes; the first scanning road segment is determined according to the road node corresponding to the first location point.

For example, the road information of the road node includes the boundary position information of the corresponding road. At this time, the area of the road corresponding to each road node can be determined from the road information of each road node. In this way, the area including the first location point can be filtered from the determined area, and the road node corresponding to the filtered area is the road node corresponding to the first location point.

In addition, after the road node corresponding to the first position point is determined, the road section on the road node corresponding to the first position point within a specified distance before and after the first position point may be used as the first scanning road section. The specified distance is pre-configured. For example, the designated distance may be 20 meters. In this case, a road section within a range of 20 meters before and after the first position point on the road node corresponding to the first position point may be used as the first scanning road section.

Optionally, after the road node corresponding to the first location point is determined, the road node corresponding to the first location point in the road topology model and other road nodes that have a connection relationship with the road node can also be used as the first scan Road section. For example, FIG. 7 is a schematic diagram of a road topology model provided by an embodiment of the present application. As shown in Fig. 7, the road topology model includes 7 road nodes, and the connection relationship between each road node is shown in Fig. 7 as the connection relationship between road nodes. That is, the road indicated by road node 1 is connected with the road indicated by road node 2, the road indicated by road node 2 is connected with the road indicated by road node 3, and the road indicated by road node 3 is connected with road node 4. The indicated roads are connected, and the roads indicated by the road node 4 are connected to the roads indicated by the road node 5, the road node 6, and the road node 7, respectively. Assuming that the road node corresponding to the first location point determined above is road node 2, at this time, the three road sections indicated by the road node 1, road node 2, and road node 3 can be taken as the first scanned road section . In addition, the road nodes in FIG. 7 that have a connection relationship with multiple road nodes may also be referred to as a junction road node (junction).

The foregoing is only an example of two possible implementations of determining the first scanning road segment according to the road node corresponding to the first location point. The embodiment of the present application does not limit the specific implementation of determining the first scanning road section according to the road node corresponding to the first location point, only the determined first scanning road section is the road section near the first location point when the autonomous vehicle is at the first location point. . This will not give examples of other implementations one by one.

After the first scanning road segment is determined through step 601, in order to facilitate the rapid determination of virtual objects distributed on the first scanning road segment, in addition to the road information of each road node, the pre-configured road topology model may also include each road The virtual object information of the node. The virtual object information of each road node includes the three-dimensional model data of each virtual object distributed in the corresponding road node. Therefore, based on step 602, the virtual objects distributed in the first scanning section can be quickly determined directly through the road topology model. It should be noted that, in the embodiment of the present application, the virtual objects distributed on the road section include virtual objects distributed on the road of the road section and on both sides of the road.

In a possible implementation manner, the three-dimensional model data of the virtual object may include three-dimensional position information of each surface point on the virtual object, so as to facilitate subsequent determination of point cloud data based on the three-dimensional position information of the surface point.

In the embodiment of the present application, the three-dimensional model data of the virtual object may be pre-processed in advance according to the requirements of the road conditions during the driving of the autonomous vehicle, so as to further increase the rate of determining the point cloud data in the future. The preprocessing methods of various three-dimensional model data provided in the embodiments of the present application will be explained below first.

In a possible implementation manner, considering that the automatic driving vehicle is driving, the detailed three-dimensional structure of nearby objects needs to be clarified to avoid traffic accidents. For distant objects, you only need to know the approximate outline to guide the current planning. Therefore, for the first virtual object among the multiple virtual objects, the three-dimensional model data of the first virtual object may also include the three-dimensional position information of the bounding box of the first virtual object. The bounding box of the first virtual object refers to the bounding box surrounding the first virtual object. The geometry of the virtual object. The first virtual object is any one of a plurality of virtual objects. That is, in the road topology model, the three-dimensional model data of each virtual object may also include the three-dimensional position information of the bounding box of the corresponding virtual object. In order to quickly determine the point cloud data according to the outline of the virtual object.

FIG. 8 is a schematic diagram of a bounding box of a virtual object provided by an embodiment of the present application. The bounding box shown in FIG. 8 can also be referred to as an AABB box. As shown in Figure 8, the AABB box is a bounding box of a three-dimensional object. The bounding box is a simple geometric space that contains virtual objects with complex shapes. The purpose of adding bounding boxes to virtual objects is to quickly perform collision detection or to filter before accurate collision detection. As shown in Figure 8, when the laser beam collides with the bounding box, the laser beam may collide with the virtual object in the bounding box. When the laser beam does not collide with the bounding box, it is impossible for the laser beam to collide with the virtual object in the bounding box. Therefore, the bounding box of the virtual object can represent the approximate outline of the virtual object.

In another possible implementation manner, considering that during the driving of the autonomous vehicle, virtual objects on both sides of the road that are particularly far away from the autonomous vehicle will not have any impact on the driving of the autonomous vehicle. Therefore, it is also possible to filter virtual objects that are particularly far apart on both sides of the road. At this time, in the road topology model, for the first road node among the multiple road nodes, the distance between the virtual object distributed on the first road node and the center line of the first road node is within the first distance threshold. That is, in the road topology model, the distance between the virtual objects on both sides of the road corresponding to each road node and the center line of the corresponding road node is within the first distance threshold.

The first distance threshold is a preset distance threshold. For example, the first distance threshold may be 10 meters. For the convenience of subsequent description, this preprocessing method is referred to as distant object removal processing. In addition, the first distance threshold also needs to be greater than half of the road width. The road width refers to the distance between the two borders on both sides of the road node.

In another possible implementation manner, considering that during the driving of the autonomous vehicle, the road and virtual objects on both sides of the road that are particularly high from the ground will not have any impact on the driving of the autonomous vehicle. For example, the high-rise buildings on both sides of the road will not have any impact on the driving of autonomous vehicles. Therefore, it is also possible to filter the road and the virtual objects on both sides of the road that are particularly high from the ground or the parts on the virtual objects. At this time, in the road topology model, the three-dimensional model data of each virtual object includes the three-dimensional position information of each of the multiple surface points of the corresponding virtual object. The three-dimensional position information includes the height of the surface points and is distributed at the road nodes. The height in the three-dimensional model data of the virtual virtual body is within the height threshold. Fig. 9 is a schematic diagram of a high-rise building provided by an embodiment of the present application. As shown in Figure 9, the three-dimensional model data of the high-rise part of the building will not be included in the road topology model. For the convenience of subsequent description, this pre-processing method is referred to as higher object removal processing.

It should be noted that the above-mentioned filtering of the three-dimensional model data of each virtual object in the road topology model can be performed in advance. At this time, the amount of data included in the road topology model is also relatively small. The road topology model directly determines the three-dimensional model data of the virtual object. Optionally, the road topology model may also only include the three-dimensional position information of each surface point on the virtual object. When the first scanning road section is determined subsequently, the three-dimensional model data of the virtual object on the first scanning road section is determined by the above-mentioned filtering method. It is enough to filter, but at this time, the three-dimensional model data needs to be filtered at any position point, and the required calculation amount is still relatively large.

In addition, considering that different processing methods can be used for virtual objects that are different from the first location point in the future, in another possible implementation, in the road topology model, the three-dimensional model data of each virtual object is also It can include the specific location of the corresponding virtual object on the road. Specifically, for any road node, a road starting point can be set on the road indicated by the road node, and the road starting point can be a position point at the center of the road at the starting position of the road. Then, for any virtual object on the road and on both sides of the road indicated by the road node, two parameters can be used to indicate the specific position of the virtual object on the road. One parameter is the distance between the center point of the virtual object's bounding box and the center line of the road, which can be marked as t-offset, and the other parameter is the center point of the virtual object's bounding box and the starting point of the road along the road driving direction The distance between can be marked as s-offset.

FIG. 10 is a schematic diagram of a road node provided by an embodiment of the present application. As shown in FIG. 10, the starting point of the road on the road indicated by the road node is point A, the t-offset of a tree in the road is 5 meters, and the s-offset is 50 meters. After the specific location of the starting point of the road is known, the distance between the tree and the first location point can be determined according to the first location point and the two parameters t-offset and s-offset.

Optionally, the specific position of each virtual object on the road in the road topology model can also be directly indicated by the absolute coordinates of the center point of the virtual object's bounding box in the geodetic coordinate system, so that for any road node, there is no need to configure The starting point of the above road.

Step 602: Determine the three-dimensional model data of each virtual object among the multiple virtual objects distributed on the first scanning road section.

Based on the explanation of the road topology model in step 601, in a possible implementation, the road topology model may include multiple road nodes, and virtual object information of each road node, and virtual object information of each road node. Including the three-dimensional model data of each virtual object distributed at the corresponding road node. At this time, the implementation of step 602 may be: obtaining three-dimensional model data of each virtual object among the multiple virtual objects distributed on the first scanning road segment from the virtual object information of each road node.

Based on the introduction of the road topology model in step 601, in a possible implementation manner, for the first virtual object among the multiple virtual objects distributed in the first scanning section, the three-dimensional model data of the first virtual object includes : The three-dimensional position information of each surface point on the first virtual object. The first virtual object is any one of the plurality of virtual objects.

In another possible implementation manner, the three-dimensional model data of the first virtual object may further include: three-dimensional position information of the bounding box of the first virtual object.

In addition, based on the above-mentioned preprocessing method for the three-dimensional model data in the road topology model, it can be known that if the three-dimensional model data of the virtual object in the road topology model is preprocessed, then what is obtained in step 602 is the preprocessed 3D model data of virtual objects.

Therefore, in a possible implementation manner, if the virtual objects in the road topology model are removed in advance, at this time, the virtual objects distributed on the first scanning section and the center line of the first scanning section are different. The distance between is within the first distance threshold. The first distance threshold will not be described in detail here.

In another possible implementation, if the virtual objects in the road topology model are processed in advance for higher object removal processing, at this time, the heights in the three-dimensional model data of the virtual objects distributed in the first scan section are all at the height Within the threshold.

In addition, in a possible implementation manner, when obtaining the three-dimensional model data of each virtual object among the multiple virtual objects distributed in the first scanning section from the virtual object information of each road node, the road topology model can be directly obtained The three-dimensional model data of the virtual object related to the first scanned section of the road is sufficient. In this case, it is assumed that when the vehicle is at the first position point, the displacement of the vehicle during the horizontal direction of the lidar is basically zero.

In another possible implementation, if the vehicle is at the first position, the displacement of the vehicle is relatively large when the lidar scans one circle in the horizontal direction. In this case, as the lidar scans, the vehicle is moving forward. Mobile. At this time, in order to improve the consistency between the point cloud data obtained by the lidar during the simulation test and the point cloud data obtained in the real world, a sub-regional method can be used to determine the multiple virtual objects distributed in the first scanning section Three-dimensional model data of each virtual object.

The implementation of determining the three-dimensional model data of each virtual object among the multiple virtual objects distributed in the first scanning section in the manner of subregions may be: dividing the first scanning section into multiple fan-shaped regions with the first position point as the center , Get the boundary information of each fan-shaped area; arrange the multiple fan-shaped areas in the order of the scanning direction of the laser detection device; for the first fan-shaped area after the arrangement, according to the boundary information of the first fan-shaped area, from each road node Obtain the three-dimensional model data of the virtual object located in the first fan-shaped area from the virtual object position information; for the i-th fan-shaped area after the arrangement, based on the scanning speed of the laser detection device and the moving speed of the vehicle, determine the laser detection device from the first The movement displacement of the vehicle in the process of scanning from a sector area to the i-th sector area, i is a positive integer greater than or equal to 2 and less than or equal to the number of divided sector areas; the i-th sector area is updated according to the movement displacement According to the updated boundary information of the i-th sector area, the three-dimensional model data of the virtual object located in the first sector area is obtained from the position information of the virtual object of each road node.

The boundary information of each sector area is used to indicate the coverage area of the corresponding sector area. For the first sector area after the arrangement, this is equivalent to the area scanned by the lidar when the vehicle is just at the first position point. At this time, the first sector area can be obtained directly from the virtual object information of each road node 3D model data of the virtual object. For the i-th sector area after the arrangement, when the lidar scans to the i-th sector area, at this time, the vehicle has moved forward by a certain displacement. Therefore, the i-th sector area can be updated according to the position of the vehicle after it has moved The boundary information. It is equivalent to dynamically updating the virtual objects in each fan-shaped area, so that the obtained virtual objects are as consistent as possible with the objects obtained in the real world.

In addition, according to the movement displacement, the realization of updating the boundary information of the i-th sector area can be: according to the first position point and the movement displacement, determine the position point of the vehicle after the movement, and divide the position point after the vehicle movement as the center. Sector regions, the multiple sector regions are arranged in the same manner as described above, and the boundary information of the i-th sector region after the arrangement is the boundary information of the i-th sector region after the update.

FIG. 11 is a schematic diagram of dividing a sector area according to an embodiment of the present application. As shown in Fig. 11, when the vehicle is at the first position point, the first scan road section can be divided into 8 fan-shaped areas with the first position point as the center. In Fig. 11, these 8 fan-shaped areas are marked as ①②③④⑤⑥⑦⑧. Suppose that when scanning to the second sector area, the vehicle moves to another location. At this time, the updated second sector area is the sector area ② shown on the right side of Fig. 11. That is, the boundary information of the right fan-shaped area ② is the updated boundary information of the second fan-shaped area.

In addition, the above-mentioned multiple fan-shaped regions can be arranged in order according to the scanning direction of the laser detection device. As to which of the multiple fan-shaped regions is the first fan-shaped region, the embodiment of the present application is not limited. For example, for the 8 fan-shaped areas shown in Figure 11, assuming that the scanning direction of the lidar is sort ①→②→③, then the order of these 8 fan-shaped areas can be fan-shaped area ①, fan-shaped area ②, fan-shaped area ③, fan-shaped Area ④, sector area ⑤, sector area ⑥, sector area ⑦, sector area ⑧. The order of these 8 sector areas can also be sector area ②, sector area ③, sector area ④, sector area ⑤, sector area ⑥, sector area ⑦, sector area ⑧, sector area ①.

In addition, the number of divided sector areas can be set according to the scanning speed of the lidar and the moving speed of the vehicle. In a possible implementation manner, the number of divided fan-shaped regions may have a negative correlation with the scanning speed of the lidar and the moving speed of the vehicle. In other words, the faster the scanning speed of the lidar, the faster the moving speed of the vehicle. At this time, the movement displacement of the vehicle during the lidar scan for one round is smaller, and the objects around the vehicle change more slowly during the lidar scan. Therefore, the number of divided fan-shaped regions can be smaller, and correspondingly, the coverage area of each fan-shaped region will be larger.

In addition, after the fan-shaped regions are divided, the process of obtaining the three-dimensional model data of the virtual objects located in each fan-shaped region can be processed in parallel, thereby improving the efficiency of finally obtaining point cloud data.

Step 603: Determine the point cloud data acquired by the laser detection device at the first position point according to the laser scanning range of the laser detection device simulated on the vehicle and the three-dimensional model data of each virtual object among the multiple virtual objects.

In the embodiment of the present application, acquiring the point cloud data acquired by the laser detection device at the first position point is essentially determining the intersection point of each laser beam on the virtual object. Therefore, in a possible implementation, step 603 may specifically be: for the first virtual object, according to the three-dimensional position information of the bounding box of the first virtual object, the direction of the vehicle, and the first position point, from the laser scanning range It is determined that the laser beam emitted by the laser detection device covers the angular range of the first virtual object. The first virtual object refers to any virtual object among multiple virtual objects; for the first laser beam in the angular range, according to the first virtual object The three-dimensional model data of determines the intersection point of the first laser beam on the first virtual object, and uses the three-dimensional position information of the intersection point as the point cloud data corresponding to the first laser beam, and the first laser beam is any laser beam in the angular range.

The foregoing first virtual object refers to any virtual object among multiple virtual objects, that is, for any virtual object determined in step 602, the foregoing implementation manner is used to determine the point where the laser beam is projected on the virtual object. Cloud data.

FIG. 12 is a schematic diagram of an angle range provided by an embodiment of the present application. As shown in Figure 12, suppose there is a target vehicle in front of the autonomous vehicle. Then, according to the orientation of the autonomous vehicle (for example, the heading of the car), the position information of the lidar (that is, the above-mentioned first position point), and the information of the AABB box of the target vehicle, the laser beam irradiated by the lidar on the target vehicle can be determined The angle range of is: horizontal a°-b°, vertical m°-n°. (In Figure 12, only the horizontal angle range is illustrated, and the vertical angle range is not shown)

Taking into account that during the driving of an autonomous vehicle, it is necessary to clarify the detailed three-dimensional structure of nearby objects to avoid traffic accidents. For distant objects, you only need to know the approximate outline to guide the current planning. Therefore, for the first laser beam in the angular range, the intersection point of the first laser beam on the first virtual object is determined according to the three-dimensional model data of the first virtual object, and different virtual objects with different distances from the first position point can be used. To determine the intersection point.

In a possible implementation manner, in the case that the first virtual object is a short-distance virtual object, the foregoing determination of the intersection point of the first laser beam on the first virtual object according to the three-dimensional model data of the first virtual object may specifically be : According to the three-dimensional position information of each surface point in the three-dimensional model data of the first virtual object, the multiple first bins of the first virtual object are determined, and each first bin refers to one of the surface points of the first virtual object. A surface formed by three or more surface points. A short-distance virtual object refers to a virtual object whose distance from the first position point is within the second distance threshold; the The first face element where the laser beams intersect, and the intersection point between the first laser beam and the first face element that intersects is taken as the intersection point of the first laser beam on the first virtual object.

That is, for a short-distance virtual object, when determining the point cloud data, the precise surface structure of the virtual object is considered.

For a short-distance virtual object, within a certain angle range, obtain point cloud data for all laser beams, because the distance and angle of each laser beam relative to the first virtual object, as well as the relative facets on the first virtual object The position is known. Therefore, you can directly locate which face element the laser beam intersects with, and then determine the intersection point detected by the laser beam through the intersection process. The position information of the intersection point is the point cloud data.

As shown in Figure 12, assuming that the first virtual object is a target vehicle opposite the vehicle, the process of obtaining point cloud data for all laser beams can be as follows:

First, take the horizontal a°, the longitudinal (m+n)/2° laser beam, and the outermost surface element A of the target vehicle (that is, the surface element where the origin in Figure 13 is located), and calculate the intersection. Get point cloud data.

Then take the horizontal (a+0.45)° ray, (0.45° is the horizontal resolution of the laser detection device) to calculate the intersection with panel A, if they intersect, get the point cloud data, if you don’t want to intersect, then take the panel The left side face element of A is intersected until the intersecting face element is found, and the point cloud data is obtained.

The calculation in the vertical direction is the same, as long as the upper and lower positions are obtained for the intersection calculation. By calculating the intersection of all laser rays in the range of a°–b° in the horizontal direction and m°-n° in the longitudinal direction, all the point cloud data of the target vehicle detected by the lidar can be obtained.

The intersection process of the above-mentioned laser beams can be processed in parallel, which will not be described in detail here.

The above is used to explain the process of obtaining point cloud data for a short-distance virtual object. In another possible implementation manner, when the first virtual object is a long-distance virtual object, the foregoing determination of the intersection point of the first laser beam on the first virtual object according to the three-dimensional model data of the first virtual object may specifically be It is: in the case that the first virtual object is a long-distance virtual object, according to the three-dimensional position information of the bounding box of the first virtual object, a plurality of second bins of the bounding box are determined, and the long-distance virtual object refers to the first position For virtual objects whose distance between the points is between the second distance threshold and the first distance threshold, the first distance threshold is greater than the second distance threshold; determine the second bin that intersects the first laser beam among the plurality of second bins , The intersection point between the first laser beam and the second intersecting surface element is taken as the intersection point of the first laser beam on the first virtual object.

That is, for a long-distance virtual object, when determining the point cloud data, the approximate outline of the virtual object is considered. For example, for the outermost buildings and walls of the road, the laser beam can be directly used to calculate the intersection with the AABB box to reduce the amount of calculation. The specific process of requesting an appointment will not be explained in detail here.

In addition, considering the occlusion between virtual objects, the same laser beam may theoretically be projected on different virtual objects, but in reality, only one point cloud data can be collected for each laser beam. For example, the angle range of the target vehicle in Figure 12 is 20°-60° horizontally and -20°-10° longitudinally, and the angle range of the laser beam projected to the wall behind the target vehicle also includes 20°-60° horizontally. , The longitudinal angle range of -20°-10°, the lateral angle range of 20°-60°, and the longitudinal angle range of -20°-10°, the laser beam only needs to perform an intersection calculation with the target vehicle, without the need for the back wall Perform repeated calculations.

Therefore, before determining the intersection of the first laser beam on the first virtual object according to the three-dimensional model data of the first virtual object, it is necessary to determine whether the projection direction of the first laser beam is whether there are other virtual objects that block the first virtual object. . If there is no other virtual object that blocks the first virtual object in the projection direction of the first laser beam, the operation of determining the intersection point of the first laser beam on the first virtual object is performed. If there are other virtual objects that block the first virtual object in the projection direction of the first laser beam, there is no need to perform the operation of determining the intersection point of the first laser beam on the first virtual object. That is, in this case, there is no need to determine the point cloud data of the projection point of the first laser beam on the first virtual object.

The foregoing determination of whether the projection direction of the first laser beam is whether there are other virtual objects that block the first virtual object may be based on the angle of the first laser beam, the first position point, and the virtual object information of each virtual object determined in step 602. Sure. In a possible implementation manner, in the projection direction of the first laser beam, if the coverage area of the bounding box of other virtual objects exists between the first position point and the first virtual virtual body, it is considered that the first laser beam There are other virtual objects that block the first virtual object in the projection direction of.

In addition, in the above process of judging whether the projection direction of the first laser beam is the presence of other virtual objects that block the first virtual object, the other virtual objects are treated as simple models of impermeable laser beams. In the real world, for objects with complex models such as trees, there are through holes in the trees that can pass the laser beam. Therefore, even if the bounding box corresponding to the tree blocks the first laser beam, the laser beam will still be projected Go to the first virtual virtual body.

Therefore, in the embodiment of the present application, if there are other virtual objects that block the first virtual object in the projection direction of the first laser beam, and the other virtual objects are non-strictly occluded objects, then it is executed to determine that the first laser beam is in the first laser beam. The operation of the intersection point on a virtual object. The non-strictly occluded object refers to an object with a through hole through which the laser beam can penetrate. At this time, after determining the intersection point of the first laser beam on the first virtual object, if the first laser beam has no intersection point on other virtual objects, set the first laser beam to the intersection point on the first virtual object. The three-dimensional position information is used as the point cloud data corresponding to the first laser beam.

If the first laser beam does not intersect on other virtual objects, it indicates that the first laser beam can pass through the through holes on other virtual objects. In this case, it is necessary to use the three-dimensional position information of the intersection point of the first laser beam on the first virtual object as the point cloud data corresponding to the first laser beam.

It should be noted that when applying the method provided in this application, the occluded virtual object can be processed without distinguishing whether it is a non-strict occlusion object, and a simple model can be used for processing, which can further reduce the amount of calculation.

Optionally, the occlusion situation may not be considered, and the point cloud data of each laser beam within the angle range may be directly determined for all virtual objects. Subsequent to the same laser beam, if the laser beam corresponds to multiple point cloud data, it indicates that the laser beam can be projected to multiple location points. At this time, it is sufficient to select the point cloud data of the position point closest to the first position point as the point cloud data of the laser beam. In this way, although the amount of calculation will be more, it can simplify the process of obtaining point cloud data.

The above steps 601 to 603 are used to explain how to determine the point cloud data acquired by the laser detection device at the first position point. In the simulation test, as the vehicle moves in the virtual space, it is necessary to update the point cloud data at each location point in time. In the process of vehicle movement, there may be some objects around the vehicle and the relative position between the vehicle has not changed. In this case, for these virtual objects, there is no need to determine the point cloud data again. Therefore, in the embodiment of the present application, the amount of calculation required to determine the point cloud data can be reduced according to the relative state between the vehicle and the surrounding virtual objects.

At this time, after the point cloud data acquired by the laser detection device at the first position point is determined in step 603, the point cloud data corresponding to each laser beam may also be cached. In the case that the vehicle is at the second location point in the virtual space, determine the second scan section according to the second location point and the road topology model; determine multiple virtual objects distributed on the second scan section; if distributed on the first scan section If the same virtual object exists in the multiple virtual objects and multiple virtual objects distributed on the second scanning section, and the relative distance between the same virtual object and the vehicle does not change at the first position point and the second position point, then The point cloud data corresponding to the laser beam projected on the same virtual object in the buffer is used as the point cloud data corresponding to the corresponding laser beam at the second position point.

The relative distance between the same virtual object and the vehicle does not change between the first position point and the second position point, indicating that the same virtual object and the vehicle remain relatively stationary during the movement of the vehicle. Therefore, there is no It is necessary to repeatedly determine the point cloud data for the same virtual object.

It should be noted that, when the vehicle is moving, if the vehicle is in a stationary state, the dynamic objects around the vehicle, such as other vehicles, will change. Therefore, the above-mentioned second location point and the first location point may be the same location point. In this case, there is no need to repeatedly calculate the point cloud data for static objects around the vehicle.

In addition, during the driving of the vehicle, some laser beams may be projected on the ground or on certain parts of the vehicle (such as the base of the lidar installed on the vehicle), and the point cloud data of these laser beams will hardly change. Therefore, in the embodiment of the present application, the point cloud data of all laser beams of the laser detection device may be initialized first. Subsequent only needs to update the initialized point cloud data according to the changed point cloud data to further reduce the amount of calculation.

As shown in Figure 14, a part of the laser beam of the lidar will hit the ground, and the point cloud data of this part will basically not change in the future; if the remaining laser does not detect the target object, the point cloud data can be set to a null value . These point cloud data can be saved first in the form of configuration files. When the simulation starts, load the configuration file to generate a point cloud cache. After the simulation starts, if the lidar detects a virtual object, the point cloud data with a null value in the cache is updated.

In addition, based on the above configuration file, if there are no other virtual objects around the vehicle within a certain period of time during the driving of the vehicle, the above configuration file can be reloaded, which is equivalent to initializing the point cloud data of all laser beams. Then continue to update the point cloud data according to the detected virtual objects.

In the embodiment of the present application, considering that the autonomous vehicle only needs to determine objects near the traveling road during the driving process, in order to reduce the amount of data processing in the process of determining the point cloud data. For the virtual space to be simulated, a road topology model for the virtual space can be constructed in advance. The road topology model is used to indicate the road topology in the virtual space. In this way, when subsequently determining the point cloud data when the vehicle is at the first position point, the current road segment to be scanned can be directly determined based on the road topology model, without the need to traverse all nodes in the KD tree as in related technologies, thereby reducing The amount of data processing in the process of obtaining point cloud data, correspondingly, improves the efficiency of obtaining point cloud data.

FIG. 15 is a schematic diagram of an apparatus for obtaining point cloud data provided by an embodiment of the present application. As shown in FIG. 15, the device 1500 includes:

The determining module 1501 is configured to determine the first scanned road segment according to the first position point and the road topology model when the vehicle is at the first position point in the virtual space, and the road topology model is used to indicate the road topology in the virtual space. For a specific implementation manner, reference may be made to step 601 in the embodiment of FIG. 6.

The determining module 1501 is also used to determine the three-dimensional model data of each virtual object among the multiple virtual objects distributed in the first scanning section. For a specific implementation manner, reference may be made to step 602 in the embodiment of FIG. 6.

The determining module 1501 is also used to determine the point cloud acquired by the laser detection device at the first position point according to the laser scanning range of the laser detection device simulated on the vehicle and the three-dimensional model data of each virtual object in the multiple virtual objects data. For a specific implementation manner, reference may be made to step 603 in the embodiment of FIG. 6.

Optionally, the road topology model includes multiple road nodes and road information of each road node in the multiple road nodes;

Determine the module to be used for:

According to the road information of each road node, determine the road node corresponding to the first location point from the multiple road nodes;

The first scanning road segment is determined according to the road node corresponding to the first position point.

Optionally, the road topology model includes multiple road nodes and virtual object information of each of the multiple road nodes. The virtual object information of the first road node includes the three-dimensional information of each virtual object distributed on the first road node. Model data, the first road node is any one of the multiple road nodes;

Determine the module to be used for:

From the virtual object information of each road node, the three-dimensional model data of each virtual object among the multiple virtual objects distributed on the first scanning road section is obtained.

Optionally, the distance between the virtual objects distributed on the first scanning road segment and the center line of the first scanning road segment is within the first distance threshold.

Optionally, the three-dimensional model data of the first virtual object in the plurality of virtual objects includes three-dimensional position information of each surface point in the plurality of surface points of the first virtual object, and the three-dimensional position information includes the height of the surface point. The virtual object is any virtual object among multiple virtual objects;

The height in the three-dimensional model data of the virtual objects distributed in the first scanning section is within the height threshold.

Optionally, the three-dimensional model data of the first virtual object in the plurality of virtual objects includes the three-dimensional position information of the bounding box of the first virtual object. The bounding box of the first virtual object refers to the geometric body surrounding the first virtual object. The object is any virtual object among multiple virtual objects;

Determine the module to be used for:

According to the three-dimensional position information of the bounding box of the first virtual object, the orientation of the vehicle, and the first position point, determine from the laser scanning range that the laser beam emitted by the laser detection device covers the angle range of the first virtual object;

Determine the intersection point of the first laser beam on the first virtual object according to the three-dimensional model data of the first virtual object, and use the three-dimensional position information of the intersection point as the point cloud data corresponding to the first laser beam. The first laser beam has an angular range Any laser beam.

Optionally, the determining module is used to:

In the case that the first virtual object is a close-range virtual object, according to the three-dimensional position information of each surface point in the three-dimensional model data of the first virtual object, multiple first face elements of the first virtual object are determined, and each first face Meta refers to the surface formed by three or more surface points among the various surface points of the first virtual object, and the short-distance virtual object refers to the virtual object whose distance from the first position point is within the second distance threshold. object;

Determine the first surface element that intersects the first laser beam among the plurality of first surface elements, and use the intersection point between the first laser beam and the intersecting first surface element as the intersection point of the first laser beam on the first virtual object .

Optionally, the determining module is used to:

In the case that the first virtual object is a long-distance virtual object, according to the three-dimensional position information of the bounding box of the first virtual object, multiple second bins of the bounding box are determined. For virtual objects whose distance is between the second distance threshold and the first distance threshold, the first distance threshold is greater than the second distance threshold;

Determine the second surface element that intersects the first laser beam among the plurality of second surface elements, and use the intersection point between the first laser beam and the intersecting second surface element as the intersection point of the first laser beam on the first virtual object .

Optionally, the determining module is also used to:

If there is no other virtual object that blocks the first virtual object in the projection direction of the first laser beam, the operation of determining the intersection point of the first laser beam on the first virtual object is performed.

Optionally, the determining module is also used to:

If there are other virtual objects that block the first virtual object in the projection direction of the first laser beam, and the other virtual objects are non-strictly occluded objects, perform the operation of determining the intersection point of the first laser beam on the first virtual object , Non-strictly shielded objects refer to objects with through holes through which the laser beam can penetrate;

After determining the intersection point of the first laser beam on the first virtual object, the method further includes:

If the first laser beam has no intersection point on the other virtual object, the three-dimensional position information of the intersection point of the first laser beam on the first virtual object is used as the point cloud data corresponding to the first laser beam.

Optionally, the determining module is used to:

Divide the first scanning road section into multiple fan-shaped areas with the first location point as the center, and obtain boundary information of each fan-shaped area;

Arrange multiple fan-shaped areas in sequence according to the scanning direction of the laser detection device;

For the first sector area after the arrangement, the three-dimensional model data of the virtual object located in the first sector area is obtained from the position information of the virtual objects of each road node according to the boundary information of the first sector area;

For the i-th sector area after the arrangement, based on the scanning speed of the laser detection device and the moving speed of the vehicle, determine the movement displacement of the vehicle during the scanning of the laser detection device from the first sector area to the i-th sector area, i is A positive integer greater than or equal to 2 and less than or equal to the number of divided sector areas;

According to the movement displacement, update the boundary information of the i-th sector area;

According to the updated boundary information of the i-th sector area, the three-dimensional model data of the virtual object located in the first sector area is obtained from the position information of the virtual object of each road node.

Optionally, the point cloud data acquired by the laser detection device at the first position point includes point cloud data corresponding to each laser beam emitted by the laser detection device;

The device also includes:

Cache module, used to cache the point cloud data corresponding to each laser beam;

The determining module is also used for determining the second scanning road section according to the second position point and the road topology model when the vehicle is at the second position point in the virtual space; determining multiple virtual objects distributed on the second scanning road section; if The same virtual object exists in the multiple virtual objects distributed on the first scanning road segment and the multiple virtual objects distributed on the second scanning road segment, and the relative distance between the same virtual object and the vehicle is at the first location point and the second location point If there is no change in time, the point cloud data corresponding to the laser beam projected on the same virtual object in the buffer is used as the point cloud data corresponding to the corresponding laser beam at the second position point.

In the embodiment of the present application, considering that the autonomous vehicle only needs to determine objects near the traveling road during the driving process, in order to reduce the amount of data processing in the process of determining the point cloud data. For the virtual space to be simulated, a road topology model for the virtual space can be constructed in advance. The road topology model is used to indicate the road topology in the virtual space. In this way, when subsequently determining the point cloud data when the vehicle is at the first position point, the current road segment to be scanned can be directly determined based on the road topology model, without the need to traverse all nodes in the KD tree as in related technologies, thereby reducing The amount of data processing in the process of obtaining point cloud data, correspondingly, improves the efficiency of obtaining point cloud data.

It should be noted that when the device for acquiring point cloud data provided in the above embodiment acquires point cloud data, only the division of the above-mentioned functional modules is used as an example for illustration. In practical applications, the above-mentioned functions can be assigned to different functions as required. The function module is completed, that is, the internal structure of the device is divided into different function modules to complete all or part of the functions described above. In addition, the device for acquiring point cloud data provided in the above-mentioned embodiment and the embodiment of the method for acquiring point cloud data belong to the same structure. For the specific implementation process, please refer to the method embodiment, which will not be repeated here.

FIG. 16 is a schematic structural diagram of a computer device provided by an embodiment of the present application. The simulation platform in FIG. 5 can be implemented by the computer device shown in FIG. 16. Referring to FIG. 16, the computer device includes at least one processor 1601, a communication bus 1602, a memory 1603, and at least one communication interface 1604.

The processor 1601 may be a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling program execution of the solution of this application.

The communication bus 1602 may include a path for transferring information between the above-mentioned components.

The memory 1603 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, random access memory (RAM), or other types that can store information and instructions The dynamic storage device can also be electrically erasable programmable read-only memory (EEPROM), compact disc read-only Memory (CD-ROM) or other optical disc storage, optical disc storage (Including compact discs, laser beam discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or can be used to carry or store desired program codes in the form of instructions or data structures and can be stored by a computer Any other media taken, but not limited to this. The memory 1603 may exist independently, and is connected to the processor 1601 through a communication bus 1602. The memory 1603 may also be integrated with the processor 1601.

Among them, the memory 1603 is used to store program codes for executing the solutions of the present application, and the processor 1601 controls the execution. The processor 1601 is configured to execute program codes stored in the memory 1603. One or more software modules can be included in the program code. The simulation platform in FIG. 5 can determine the data used to develop the application through one or more software modules in the program code in the processor 1601 and the memory 1603. The one or more software modules can be any of the modules in FIG. 15.

Communication interface 1604, using any device such as a transceiver to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc. .

In a specific implementation, as an embodiment, the computer device may include multiple processors, such as the processor 1601 and the processor 1605 shown in FIG. 16. Each of these processors can be a single-CPU (single-CPU) processor or a multi-core (multi-CPU) processor. The processor here may refer to one or more devices, circuits, and/or processing cores for processing data (for example, computer program instructions).

The above-mentioned computer equipment may be a general-purpose computer equipment or a special-purpose computer equipment. In a specific implementation, the computer device may be a desktop computer, a portable computer, a network server, a PDA (personal digital assistant, PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. The embodiments of this application do not limit the type of computer equipment.

In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented by software, it can be implemented in the form of a computer program product in whole or in part. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on the computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website, computer, server, or data center. Transmission to another website, computer, server or data center via wired (for example: coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (for example: infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or a data center integrated with one or more available media. The usable medium may be a magnetic medium (for example: floppy disk, hard disk, magnetic tape), optical medium (for example: digital versatile disc (DVD)), or semiconductor medium (for example: solid state disk (SSD)) )Wait.

A person of ordinary skill in the art can understand that all or part of the steps in the above embodiments can be implemented by hardware, or by a program to instruct relevant hardware. The program can be stored in a computer-readable storage medium. The storage medium mentioned can be a read-only memory, a magnetic disk or an optical disk, etc.

The above-mentioned examples provided for this application are not intended to limit this application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included in the protection scope of this application. Inside.

Claims (15)

1. A method for obtaining point cloud data, characterized in that the method is used in a computing device, and the method includes:

   In the case that the vehicle is at the first position point in the virtual space, determine the first scanned road section according to the first position point and a road topology model, where the road topology model is used to indicate the road topology in the virtual space;

   Determining the three-dimensional model data of each virtual object among the multiple virtual objects distributed on the first scanning road section;

   Determine the point cloud acquired by the laser detection device at the first position point according to the laser scanning range of the laser detection device simulated on the vehicle and the three-dimensional model data of each virtual object in the plurality of virtual objects data.
2. The method according to claim 1, wherein the road topology model comprises a plurality of road nodes and virtual object information of each road node in the plurality of road nodes, and the first among the plurality of road nodes The virtual object information of a road node includes three-dimensional model data of each virtual object distributed on the first road node, and the first road node is any one of the multiple road nodes;

   The determining the three-dimensional model data of each virtual object among the multiple virtual objects distributed on the first scanning road section includes:

   From the virtual object information of each road node, three-dimensional model data of each virtual object among the multiple virtual objects distributed on the first scanning road section is obtained.
3. The method according to claim 2, wherein the distance between the virtual objects distributed on the first scanning road segment and the center line of the first scanning road segment is within a first distance threshold.
4. The method according to claim 2 or 3, wherein the three-dimensional model data of the first virtual object in the plurality of virtual objects includes the three-dimensional position of each surface point in the plurality of surface points of the first virtual object Information, the three-dimensional position information includes the height of a surface point, and the first virtual object is any one of the multiple virtual objects;

   The height in the three-dimensional model data of the virtual objects distributed on the first scanning section is within a height threshold.
5. The method according to any one of claims 1 to 4, wherein the three-dimensional model data of a first virtual object in the plurality of virtual objects includes three-dimensional position information of a bounding box of the first virtual object, and the first virtual object The bounding box of a virtual object refers to a geometric body surrounding the first virtual object;

   According to the laser scanning range of the laser detection device simulated on the vehicle and the three-dimensional model data of each virtual object in the plurality of virtual objects, it is determined that the laser detection device acquired at the first position point Point cloud data, including:

   According to the three-dimensional position information of the bounding box of the first virtual object, the orientation of the vehicle, and the first position point, it is determined from the laser scanning range that the laser beam emitted by the laser detection device covers the first The angular range of a virtual object;

   Determine the intersection point of the first laser beam on the first virtual object according to the three-dimensional model data of the first virtual object, and use the three-dimensional position information of the intersection point as the point cloud data corresponding to the first laser beam, The first laser beam is any laser beam in the angular range.
6. The method of claim 5, wherein the determining the intersection point of the first laser beam on the first virtual object according to the three-dimensional model data of the first virtual object comprises:

   In the case that the first virtual object is a short-distance virtual object, determine the multiple first bins of the first virtual object according to the three-dimensional position information of each surface point in the three-dimensional model data of the first virtual object , Each first facet refers to a surface formed by three or more surface points among the various surface points of the first virtual object, and the distance between the near-distance virtual object and the first position point The distance is within the second distance threshold;

   Determine the first surface element that intersects the first laser beam among the plurality of first surface elements, and use the intersection point between the first laser beam and the intersecting first surface element as the first laser beam The intersection point of the light beam on the first virtual object.
7. The method of claim 5, wherein the determining the intersection point of the first laser beam on the first virtual object according to the three-dimensional model data of the first virtual object comprises:

   In the case that the first virtual object is a remote virtual object, a plurality of second bins of the bounding box are determined according to the three-dimensional position information of the bounding box of the first virtual object, and the remote virtual object The distance from the first position point is between a second distance threshold and a first distance threshold, and the first distance threshold is greater than the second distance threshold;

   Determine the second surface element that intersects the first laser beam among the plurality of second surface elements, and use the intersection point between the first laser beam and the intersecting second surface element as the first laser beam The intersection point of the light beam on the first virtual object.
8. The method according to claim 2, wherein the determining the three-dimensional model data of each virtual object among the multiple virtual objects distributed on the first scanning road section according to the virtual object information of each road node comprises:

   Dividing the first scanning road section into a plurality of fan-shaped areas centered on the first position point, and obtaining boundary information of each fan-shaped area;

   Arranging the plurality of fan-shaped areas in order according to the scanning direction of the laser detection device;

   For the first fan-shaped area after the arrangement, according to the boundary information of the first fan-shaped area, obtain the three-dimensional model data of the virtual object located in the first fan-shaped area from the virtual object position information of each road node;

   For the i-th sector area after the arrangement, based on the scanning speed of the laser detection device and the moving speed of the vehicle, it is determined that the laser detection device scans from the first sector area to the i-th sector area The movement displacement of the vehicle in the process, where i is a positive integer greater than or equal to 2 and less than or equal to the number of divided sector regions;

   Update the boundary information of the i-th sector area according to the movement displacement;

   According to the updated boundary information of the i-th sector area, the three-dimensional model data of the virtual object located in the first sector area is obtained from the position information of the virtual object of each road node.
9. The method according to any one of claims 1 to 8, wherein the point cloud data acquired by the laser detection device at the first position point includes the point corresponding to each laser beam emitted by the laser detection device Cloud data

   After determining the point cloud data acquired by the laser detection device at the first position point, the method further includes:

   Cache the point cloud data corresponding to each laser beam;

   In a case where the vehicle is at a second position point in the virtual space, determining a second scanning road section according to the second position point and the road topology model;

   Determining multiple virtual objects distributed on the second scanning road section;

   If the same virtual object exists among the multiple virtual objects distributed on the first scanning road segment and the multiple virtual objects distributed on the second scanning road segment, and the relative distance between the same virtual object and the vehicle is between If there is no change between the first position point and the second position point, the point cloud data corresponding to the laser beam projected on the same virtual object in the buffer is used as the corresponding laser beam corresponding to the second position point Point cloud data.
10. A device for acquiring point cloud data, characterized in that the device comprises:

    The determining module is used to determine the first scanning road section according to the first position point and the road topology model when the vehicle is at the first position point in the virtual space, and the road topology model is used to indicate the position in the virtual space Road topology;

    Determining the three-dimensional model data of each virtual object among the multiple virtual objects distributed on the first scanning road section;

    Determine the point cloud acquired by the laser detection device at the first position point according to the laser scanning range of the laser detection device simulated on the vehicle and the three-dimensional model data of each virtual object in the plurality of virtual objects data.
11. The device according to claim 10, wherein the road topology model comprises a plurality of road nodes, and virtual object information of each road node in the plurality of road nodes, and the first among the plurality of road nodes The virtual object information of a road node includes three-dimensional model data of each virtual object distributed on the first road node, and the first road node is any one of the multiple road nodes;

    The determining module is used for:

    From the virtual object information of each road node, three-dimensional model data of each virtual object among the multiple virtual objects distributed on the first scanning road section is obtained.
12. The device according to claim 10 or 11, wherein the three-dimensional model data of a first virtual object in the plurality of virtual objects includes three-dimensional position information of a bounding box of the first virtual object, and the first virtual object The bounding box of an object refers to a geometric body surrounding the first virtual object, and the first virtual object is any virtual object among the multiple virtual objects;

    The determining module is used for:

    According to the three-dimensional position information of the bounding box of the first virtual object, the orientation of the vehicle, and the first position point, it is determined from the laser scanning range that the laser beam emitted by the laser detection device covers the first The angular range of a virtual object;

    Determine the intersection point of the first laser beam on the first virtual object according to the three-dimensional model data of the first virtual object, and use the three-dimensional position information of the intersection point as the point cloud data corresponding to the first laser beam, The first laser beam is any laser beam in the angular range.
13. The device according to claim 11, wherein the determining module is configured to:

    Dividing the first scanning road section into a plurality of fan-shaped areas centered on the first position point, and obtaining boundary information of each fan-shaped area;

    Arranging the plurality of fan-shaped areas in order according to the scanning direction of the laser detection device;

    For the first fan-shaped area after the arrangement, according to the boundary information of the first fan-shaped area, obtain the three-dimensional model data of the virtual object located in the first fan-shaped area from the virtual object position information of each road node;

    For the i-th sector area after the arrangement, based on the scanning speed of the laser detection device and the moving speed of the vehicle, it is determined that the laser detection device scans from the first sector area to the i-th sector area The moving displacement of the vehicle in the process, where i is a positive integer greater than or equal to 2 and less than or equal to the number of divided sector regions;

    Update the boundary information of the i-th sector area according to the movement displacement;

    According to the updated boundary information of the i-th sector area, the three-dimensional model data of the virtual object located in the first sector area is obtained from the position information of the virtual object of each road node.
14. The device according to any one of claims 10 to 13, wherein the point cloud data acquired by the laser detection device at the first position point includes the point corresponding to each laser beam emitted by the laser detection device Cloud data

    The device also includes:

    Cache module, used to cache the point cloud data corresponding to each laser beam;

    The determining module is further configured to determine a second scanning road section according to the second position point and the road topology model when the vehicle is at the second position point in the virtual space; 2. Multiple virtual objects on the scanning road section; if the same virtual object exists among the multiple virtual objects distributed on the first scanning road section and the multiple virtual objects distributed on the second scanning road section, and the same virtual object and If the relative distance between the vehicles does not change between the first position point and the second position point, the point cloud data corresponding to the laser beam projected on the same virtual object in the buffer is used as the corresponding laser beam Point cloud data corresponding to the second location point.
15. A computing device, wherein the computing device includes a memory and a processor, the memory is used to store computer instructions, and the processor is used to read the computer instructions to execute any one of claims 1-9 The method described.

Images