JP4333611B2 - Obstacle detection device for moving objects - Google Patents

Description

  The present invention relates to an obstacle detection apparatus for a moving body that determines the presence or absence of an obstacle using a distance image while the mobile body is traveling.

  In general, a technique is known in which a presence image of an object such as an intruder is extracted from a distance image using a difference image between a target distance image and a background distance image (standard pattern). (For example, refer to Patent Document 1).

In addition, in order to detect obstacles on the floor where the robot walks, the Hough transform technology is applied to images with 3D information. Techniques for extraction have been proposed (see, for example, Non-Patent Document 1).
Japanese Patent Publication No. 8-21149 Kotaro Sabe et al., “Obstacle Avoidance and Walking Planning Using Stereo Images by Biped Robot”, Proceedings of 8th Image Sensing Symposium, p. 237-242, 2002

  The technique described in Patent Document 1 can be used when the distance image sensor is fixed at a fixed position. However, when this technique is applied to a moving body, a distance image that becomes a background due to vibration during traveling is used. And a positional relationship between the target distance image and the target distance image. Therefore, a mechanism such as a posture stabilization device is required so that the distance image sensor is not affected by vibrations during traveling of the moving body.

  On the other hand, since the technique described in Non-Patent Document 1 detects the floor surface based on information included in the parallax image, a mechanism such as a posture stabilization device is unnecessary, but it is necessary to perform Hough transform or the like. There is a problem that an expensive processor having a high processing capability is required to determine the presence or absence of an obstacle during traveling of the moving body due to a large load.

  The present invention has been made in view of the above-mentioned reasons, and an object of the present invention is to make it possible to easily extract a surface on which a moving body travels in a distance image using information obtained from the distance image, and thereby to stabilize the posture. It is an object of the present invention to provide an obstacle detection device for a moving body that eliminates the need for such a mechanism and can reduce the processing load as compared with the Hough transform.

  The invention according to claim 1 is a distance image sensor that is mounted on a movable body main body and captures a spatial region including a surface on which the movable body main body is going to run obliquely downward, and generates a distance image whose pixel value is a distance value; A reference image storage unit that stores a distance image of a plane within an angle range that is predicted to be generated by the distance image sensor when the body body travels on a plane as a candidate reference image at a predetermined step size, and a reference image A candidate reference image having a maximum area of a matching region whose difference value is equal to or less than a predetermined threshold in a difference image from a distance image generated by a distance image sensor among a plurality of candidate reference images stored in a storage unit is a reference distance. A reference distance image selection unit that selects as an image, and a determination unit that determines an area other than the matching area in the difference image with respect to the reference distance image selected by the reference distance image selection unit as an area where an obstacle exists. The features.

  According to this configuration, a distance image of a plane within an angle range predicted to be generated by the distance image sensor is prepared as a candidate reference image with a predetermined step size, and the distance image generated by the distance image sensor is prepared. Since the candidate reference image having the largest matching area when the difference image is generated is adopted as the reference distance image, the reference distance image corresponding to the posture of the distance image sensor when the distance image is generated is obstruction. It can be used for evaluation. Therefore, there is no need to consider the attitude of the distance image sensor, and there is a mechanism for stabilizing the attitude of the distance image sensor and the processing load for calculating a reference distance image for comparison with the distance image according to the attitude of the distance image sensor. High calculation processing becomes unnecessary.

  According to a second aspect of the present invention, in the first aspect of the present invention, in the difference image between the default candidate reference image stored in the reference image storage unit and the distance image generated by the distance image sensor, a difference value is a prescribed threshold value. A difference image generation unit having a function for obtaining a matching area as described below and a direction orthogonal to the main axis of the matching area are used to estimate the inclination direction of the traveling surface of the moving body, and within a predetermined range with respect to the inclination direction. A candidate image narrowing unit that extracts the candidate reference image from the reference image storage unit, and the reference distance image selecting unit selects a reference distance image from the candidate reference images narrowed down by the candidate image narrowing unit. To do.

  According to this configuration, since the range of the candidate reference image corresponding to the distance image can be narrowed down using the difference image between the default candidate reference image and the actual distance image, a large number of images stored in the reference image storage unit are stored. The area of the coincidence area may be evaluated for a small number of candidate reference images narrowed down from the candidate reference images. As a result, the processing load is reduced and high-speed processing can be expected.

  According to a third aspect of the present invention, in the first aspect of the present invention, in the difference image between the default candidate reference image stored in the reference image storage unit and the distance image generated by the distance image sensor, a difference value is a prescribed threshold value. A difference image generation unit having a function for obtaining a matching area as described below and a width of the matching area in a direction orthogonal to the main axis of the matching area are used to estimate the inclination angle of the traveling surface of the moving body, and to the inclination angle And a candidate image narrowing-down unit that extracts a candidate reference image within a predetermined range from the reference image storage unit, and the reference distance image selecting unit selects a reference distance image from the candidate reference images narrowed down by the candidate image narrowing-down unit It is characterized by doing.

  According to this configuration, since the range of the candidate reference image corresponding to the distance image can be narrowed down using the difference image between the default candidate reference image and the actual distance image, a large number of images stored in the reference image storage unit are stored. The area of the coincidence area may be evaluated for a small number of candidate reference images narrowed down from the candidate reference images. As a result, the processing load is reduced and high-speed processing can be expected.

  According to a fourth aspect of the present invention, in the first aspect of the present invention, in the difference image between the default candidate reference image stored in the reference image storage unit and the distance image generated by the distance image sensor, a difference value is a prescribed threshold value. A difference image generation unit having a function for obtaining a matching area, and a direction orthogonal to the main axis of the matching area, and a direction orthogonal to the main axis of the matching area while estimating the inclination direction of the traveling surface of the moving body. The candidate image is narrowed down by estimating the inclination angle of the traveling surface of the moving body using the width of the coincidence area and extracting candidate reference images within a predetermined range with respect to the combination of the inclination direction and the inclination angle from the reference image storage unit And the reference distance image selection unit selects a reference distance image from the candidate reference images narrowed down by the candidate image narrowing unit.

  According to this configuration, the range of the candidate reference image corresponding to the distance image is narrowed down using the difference image between the default candidate reference image and the actual distance image, and both the inclination direction and the inclination angle of the traveling surface are used. Since the range of candidate reference images is narrowed, it is only necessary to evaluate the area of the matching region for a small number of candidate reference images narrowed down from a large number of candidate reference images stored in the reference image storage unit, resulting in a reduction in processing load. Thus, processing at a higher speed than that of the second and third aspects can be expected.

  According to a fifth aspect of the invention, in the first aspect of the invention, the reference distance image selecting unit selects a candidate reference image within a range defined for a reference distance image used last time with respect to a distance image obtained as a time series. A reference distance image is selected.

  According to this configuration, since the next reference distance image is selected within a range that does not deviate significantly from the previously used reference distance image, it is sufficient to narrow down to a small number of candidate reference images and evaluate the area of the matching region, thereby reducing the processing load. High-speed processing can be expected.

  The invention according to claim 6 is a distance image sensor that captures a spatial region including a surface that is mounted on a mobile body and that the mobile body is about to travel, and generates a distance image having a pixel value as a distance value. A reference image storage unit that stores a distance image of a plane within an angle range that is predicted to be generated by the distance image sensor when the body body travels on a plane as a candidate reference image at a predetermined step size, and a reference image A candidate reference image having a maximum area of a matching region whose difference value is equal to or less than a predetermined threshold in a difference image from a distance image generated by a distance image sensor among a plurality of candidate reference images stored in a storage unit is a reference distance. A reference distance image selection unit that is selected as an image, and a matching area is divided into a proximity area close to the mobile body and a remote area far from the mobile body, and the area occupied by the matching area in the proximity area is less than or equal to a first threshold value If the condition that the area occupied by the coincidence area in the far region is equal to or greater than the second threshold is satisfied, the coincidence area is determined as an area where an obstacle exists, and if the condition is not met, the reference distance image selection unit And a determination unit that determines that an area other than the matching area in the difference image with respect to the reference distance image selected in step 1 is an area where an obstacle exists.

  According to this configuration, a distance image of a plane within an angle range predicted to be generated by the distance image sensor is prepared as a candidate reference image with a predetermined step size, and the distance image generated by the distance image sensor is prepared. Since the candidate reference image having the largest matching area when the difference image is generated is adopted as the reference distance image, the reference distance image corresponding to the posture of the distance image sensor when the distance image is generated is obstruction. It can be used for evaluation. Therefore, there is no need to consider the attitude of the distance image sensor, and there is a mechanism for stabilizing the attitude of the distance image sensor and the processing load for calculating a reference distance image for comparison with the distance image according to the attitude of the distance image sensor. High calculation processing becomes unnecessary. Moreover, even when the selected reference distance image matches the obstacle, the obstacle can be separated in the distance image.

  According to the configuration of the present invention, a distance image of a plane within an angle range predicted to be generated by the distance image sensor is prepared as a candidate reference image with a predetermined step size, and a distance from a plurality of candidate reference images is prepared. Since the reference distance image that is highly likely to match the reference plane of the image is selected, the reference distance image corresponding to the attitude of the distance image sensor when the distance image is generated is used for the distance image of the reference plane. There is an advantage that a mechanism for stabilizing the posture of the image sensor is not necessary, and a high processing load calculation process for obtaining a reference distance image for comparison with a distance image according to the posture of the distance image sensor is not required. .

  In embodiment described below, the autonomous mobile apparatus which moves autonomously using the map information of a movement range is illustrated. As shown in FIG. 1, the autonomous mobile device includes a mobile body 2 having a drive source such as an electric motor or an engine, and an obstacle detection device 1 mounted on the mobile body 2. As the movable body 2, a configuration is adopted in which a tire is rotated by a drive source such as an electric motor to obtain a propulsive force. For example, if the drive source is an electric motor, the mobile body 2 controls the energization current to the electric motor upon receiving an instruction regarding the propulsion force and the traveling direction from the outside, and adjusts the propulsion force by increasing or decreasing the energization current. The traveling direction is changed by controlling the energization current so as to vary the rotation speed of the tire. You may employ | adopt the structure steered so that the direction of a tire may be changed to the change of advancing direction.

  Although the configuration of the mobile body 2 is not described in detail, the mobile body 2 generally has a function of recognizing the current position, and autonomously travels so as to pass the route indicated in the map information that it holds. However, since the main part of this embodiment is in the function of detecting an obstacle to travel while avoiding the obstacle, the obstacle detection apparatus 1 will be described below.

  The obstacle detection apparatus 1 includes a distance image sensor 11 having a predetermined range including the front of the mobile body 2 as a field of view, and an image for extracting an area where an obstacle exists from the distance image generated by the distance image sensor 11. It is comprised with the image processing apparatus 10 which processes. The image processing apparatus 10 is configured by a computer that executes a program that implements the functions described below. The mobile body 2 is connected via an appropriate interface. Since the mobile body 2 generally includes a network interface for communicating with the outside, the obstacle detection device 1 can be connected to the mobile body 2 via the network interface.

  The distance image sensor 11 obtains a distance to an object existing in the field of view, and generates a distance image whose pixel value is a distance value. The principle of this type of distance image sensor is to obtain the distance to an object by stereo imaging using the parallax of multiple imaging devices, project signal light with modulated intensity, and depend on the object of the signal light. It is known that the reflected light is received and the time difference between when the signal light is projected and when it is received is obtained from the phase difference of the modulation waveform. Here, it is assumed that a configuration is employed in which signal light whose intensity is modulated is projected to obtain a phase difference between light projection and reception. An infrared light emitting diode is used for projecting signal light, and a CCD image sensor or an image sensor having a similar configuration is used for receiving signal light. In addition, a light receiving optical system including a lens group is disposed in front of the CCD image sensor.

  The distance image sensor 11 using the stereo imaging method or the phase difference between light projection and reception can measure the distance of the entire region in the field of view using information obtained within a time shorter than a fraction of a second. The distance can be measured in a shorter time than when the method is used. Therefore, if the mobile body 2 is not moving at a particularly high speed, the distance of the entire region in the field of view can be measured without a time difference even while the mobile body 2 is moving. Further, the distance image sensor 11 performs a process of converting the amount of received light into a distance in a time within one frame of the image sensor. The operation of the image processing apparatus 10 described below is also performed in the time within one frame of the image sensor. The number of frames per second is set according to the speed of the mobile body 2. In this embodiment, the speed of the mobile body 2 is assumed to be 0.5 to 1 m / s, and processing of 5 to 6 frames per second is performed. Shall be performed.

  The distance image sensor 11 is attached to the upper part of the mobile body 2 and the field of view Vf of the distance image sensor 11 is a surface (usually road surface, ground, floor surface) on which the mobile body 2 is to travel as shown in FIG. Any one of the above) is a space area in front of the mobile body 2 and is set obliquely downward. Since the distance image sensor 11 is intended to detect the presence or absence of an obstacle in the range in which the mobile body 2 travels, the viewing angle of the distance image sensor 11 is appropriately set according to the travel performance of the mobile body 2. Is done. The visual field is set wider as the speed of the mobile body 2 is higher and the turning performance is higher.

  Each pixel of the distance image sensor 11 corresponds to an angle (a dip angle and an azimuth angle) in spherical coordinates with a predetermined position of the light receiving system as an origin. That is, it is possible to specify the three-dimensional coordinate position of the object at the spherical coordinates set in the distance image sensor 11 based on the information on the angle of the spherical coordinates that the pixel position has and the information on the distance of the spherical coordinates that the pixel value has. it can.

  In order to extract the obstacle existence area from the distance image, the distance image in the state where no obstacle exists on the surface on which the mobile body 2 travels and the distance image generated from the real space by the distance image sensor 11 are used. Use the difference. When it is assumed that no obstacle exists on the surface on which the mobile body 2 travels, this surface is used as a reference surface. In other words, a plane obtained by extending the surface on which the mobile body 2 is traveling becomes the reference surface. Therefore, it is determined that there is a possibility that there is an obstacle in a region where the difference between the distance image generated by the distance image sensor 11 and the distance image of the reference plane is larger than the specified determination threshold.

  By the way, since the attitude of the distance image sensor 1 mounted on the mobile body 2 changes due to vibrations of the mobile body 2 and the like, a reference plane in which a difference image is to be generated among the distance images obtained by the distance image sensor 11. Also changes. In this embodiment, when the reference surface in a state where the mobile body 2 is stationary is the xy plane and the front of the mobile body 2 is the positive direction of the x axis, the change of the reference surface is rotated about the x axis. (Roll), and rotation (pitch) around the y-axis. In the following, in order to simplify the description, the rotation around the z-axis (yaw), the linear movement in the x-axis direction, the y-axis direction, and the z-axis direction will be described without particular consideration. Changes also include changes in these factors. This point will be described later.

  A feature of the present invention is that distance images relating to all reference planes having a predetermined step size predicted as a reference plane at the time of traveling of the mobile body 2 are prepared as candidate reference images, from among a large number of candidate reference images. An optimum reference plane image is selected as a reference distance image that is a distance image of the reference plane, and an obstacle is detected using a difference image between the selected reference distance image and the distance image generated by the distance image sensor 11. In the point.

  Therefore, in the present embodiment, for the angle change of the roll and pitch, the reference image storage unit 15 that stores all candidate reference images with a step size (angle resolution) of 1 degree within an angle range in which the change is assumed; A difference image generation unit 12 that generates a difference image between the distance image captured by the distance image sensor 11 and each candidate reference image stored in the reference image storage unit 15 is provided. For example, if the range of angles assumed to be changed is ± 10 degrees and ± 20 degrees for the roll and pitch, respectively, 20 × 40 = 800. In this case, 800 candidate reference images are stored in the reference image storage unit. 15 may be stored.

  Here, if the difference image generation unit 12 generates difference images between the distance image generated by the distance image sensor 11 and all 800 candidate reference images, the processing load becomes relatively large. Therefore, it is necessary to narrow down candidate reference images for actually generating difference images among candidate reference images stored in the reference image storage unit 15. For this purpose, a candidate image narrowing unit 16 is provided. The operation of the candidate image narrowing unit 16 will be described later.

  The candidate image narrowing unit 16 narrows down to about ten candidate reference images for generating a difference image, and gives each candidate reference image to the difference image generation unit 12 to generate a difference image between each candidate reference image and a distance image. When the difference images between the distance image and each candidate reference image are obtained, the region in which the pixel value (that is, the difference value) is close to 0 in the difference image is a region having a high degree of coincidence with the candidate reference image. I can say that. When the distance image is a reference plane that does not include an obstacle and the candidate reference image matches the reference plane of the distance image, all pixel values of the difference image are ideally 0, so the distance image has an obstacle. Even when an object is included, in the candidate reference image having the highest degree of coincidence with the reference plane included in the distance image among the plurality of candidate reference images, the area of the area where the pixel value of the difference image is close to 0 It is considered to be the maximum. Therefore, the image processing apparatus 10 obtains the area of the coincidence region where the pixel value is equal to or less than a predetermined threshold for each difference image between each candidate reference image and the distance image, and uses the candidate reference image with the maximum area as a reference. A reference distance image selection unit 13 for selecting as a distance image is provided.

  After selecting the reference distance image by the reference image selection unit 13, a normal method for detecting an obstacle can be used. Here, since the difference image between the distance image generated by the distance image sensor 11 and the reference distance image has already been generated by the difference image generation unit 13, this difference image is given to the determination unit 14. The determination unit 14 extracts a region where the pixel value exceeds the specified threshold value in the difference image as a region where an obstacle may exist. Here, the discrimination threshold is not necessarily the same as the threshold for extracting the coincidence region, but if the two match, the reference distance image selection unit 13 selects the reference distance image from the candidate reference images. It is possible to use the information obtained by binarization performed at the time as it is for evaluation of obstacles, leading to a reduction in processing load.

  By the way, the candidate image narrowing unit 16 narrows down the candidate reference images to be given to the difference image generating unit 12 from the candidate reference images stored in the reference image storage unit 15 in the following procedure. While the mobile body 2 is traveling, distance images are generated one after another as time elapses, and the mobile body 2 travels at a time interval such that a pair of distance images adjacent in time series can be obtained. It is considered that the inclination of the surface and the attitude of the distance image sensor 11 do not vary greatly. That is, in this time interval, it is considered that the change in roll and pitch is small for the reference distance image selected by the reference distance image selection unit 13. Therefore, the distance images are generated one after another with the passage of time, and the previous reference distance image can be adopted, and the roll and pitch of the candidate reference image selected as the previous reference distance image are selected. When the angle is (θ, φ), it is estimated that the current reference distance image can be selected within a specified angle range (± Δα, ± Δβ), and the candidate image narrowing unit 16 (θ ± Δα, φ ± Candidate reference images in the range of Δβ) are extracted from the reference image storage unit 15 and given to the difference image generation unit 13. Here, for example, if α = β = 3 degrees is set, the number of candidate reference images provided to the difference image generation unit 13 can be reduced to 36, and the processing load when generating the difference image can be reduced. Can do. Note that this procedure depends on the relationship between the speed of the mobile body 2 and the number of frames per second of the distance image sensor 11, and the relationship between the number of frames and the speed is the correlation between adjacent reference distance images in time series. Cannot be adopted when the relationship is such that the speed is low (a relationship in which the speed is higher than the number of frames per second).

  When the mobile body 2 is running and there is almost no change in speed or direction, the candidate reference image can be narrowed down by the above-described procedure. However, when the mobile body 2 starts to move or the speed of the mobile body 2 is When a large change occurs in the direction, the previous reference distance image cannot always be used. In short, when the correlation between adjacent reference distance images in the time series is low, the candidate reference images cannot be narrowed down using the relationship between the reference distance images. Therefore, in such a case, the following procedure is adopted. First, a candidate reference image whose roll and pitch angles are predetermined values (δ, ε) is used as a default. If δ = ε = 0, it corresponds to a distance image when the horizontal plane is used as a reference plane while the moving body 2 is stationary.

  Since the correlation between adjacent reference distance images in the time series can be evaluated according to the operation of the mobile body 2, in the case of an operation in which the correlation is equal to or less than a specified value (for example, transmission from a stopped state) ), The candidate reference images are narrowed down by the following procedure. First, the candidate image narrowing unit 16 gives the default candidate reference image stored in the reference image storage unit 15 and the distance image generated by the distance image sensor 11 to the difference image generation unit 13, and obtains the difference image between them. Generate. Except for the case where the distance plane P1 generated by the distance image sensor 11 and the default candidate reference image P0 match the reference plane, as shown in FIG. 3A, the distance image P1 is the candidate reference image P0. Is tilted back and forth (rolling), as shown in FIG. 3B, the distance image P1 is tilted left and right with respect to the candidate reference image P0 (pitching), or tilted back and forth. And the left and right slopes.

  When tilted back and forth, as shown in FIG. 4A, the matching area D1 in which the absolute value of the pixel value is equal to or less than a predetermined threshold in the difference image is horizontally long (the longitudinal direction is the y-axis direction), and left and right 4B, the matching region D1 in the difference image is vertically long (the longitudinal direction is the x-axis direction) as shown in FIG. Therefore, for the coincidence region D1, the direction of the main axis is extracted and the maximum width in the direction orthogonal to the direction of the main axis is extracted. This process is performed in the candidate image narrowing unit 16.

  Note that the default candidate reference image P0 is set so that the matching region D1 occurs near the center of the distance image. In the above example, the roll is rotated about the y-axis and the pitch is rotated about the x-axis for the sake of simplicity. However, the roll in this embodiment is not a pure roll but the x-axis direction and the z-axis. In addition, the pitch includes not only a pure pitch but also a displacement in the y-axis direction and the z-axis direction. Accordingly, the candidate reference image stored in the reference image storage unit 15 is actually corrected by appropriately including not only rotation but also displacement in the x-axis direction, the y-axis direction, and the z-axis direction.

  As apparent from a comparison between FIG. 3 and FIG. 4, the direction orthogonal to the direction of the main axis represents the inclination direction, and the maximum width in the direction orthogonal to the direction of the main axis corresponds to the inclination angle. If one of the tilt direction and the tilt angle is known, the range of combinations of the roll angle θ and the pitch angle φ corresponding to the tilt direction or tilt angle can be known. If both the tilt direction and the tilt angle are known, the combination of the roll angle θ and the pitch angle φ can be determined as one value. In the present embodiment, the angle (θ, φ) is specified by preparing a table in which the angle (θ, φ) is associated with the combination of the tilt direction and the tilt angle.

  When the angle (θ, φ) is obtained as described above, the candidate image narrowing unit 16 stores the reference image storage unit 15 in an appropriate angle range (± Δα, ± Δβ) with respect to the angle (θ, φ). The candidate reference image is extracted from the image, and a difference image between the extracted candidate reference image and the distance image generated by the distance image sensor 11 is obtained by the difference image generation unit 12. The processing after generating the difference image is the same as the processing described above, and the reference distance image selection unit 13 uses the candidate reference image that has the largest area of the matching region included in the difference image as the reference distance image. The determination unit 14 determines the presence or absence of an obstacle using the reference distance image.

  The operations of the candidate narrowing unit 16 are summarized as shown in FIG. That is, when the reference distance image used last time can be adopted (S1), the angle (θ, φ) of the previous reference distance image is obtained (S2). For this angle (θ, φ), candidate reference images in the range of (θ ± Δα, φ ± Δβ) are extracted from the reference image storage unit 15 and given to the difference image generation unit 13 (S3).

  On the other hand, when the previous reference distance image cannot be adopted (S1), a candidate reference image having a default angle (δ, ε) is selected from the reference image storage unit 15 (S4), and the distance image sensor 11 is selected. A difference image between the distance image generated in step 1 and the candidate reference image is generated (S5). A pixel area in which the absolute value of the pixel value is equal to or less than a predetermined threshold is extracted as a matching area from the difference image (S6), and a direction orthogonal to the main axis of the matching area and the maximum width of the matching area in the direction are obtained (S7) . Next, the direction orthogonal to the main axis is set as the inclination direction (gradient direction), and the inclination angle (θ, φ) is estimated from the maximum width (S8). When the angles (θ, φ) are obtained in this way, candidate reference images in the range of (θ ± Δα, φ ± Δβ) are extracted from the reference image storage unit 15 and given to the difference image generation unit 13 (S9). Instead of using the maximum width in the direction orthogonal to the main axis, it is also possible to use an average value of widths or use an average value of widths within a predetermined range with respect to the midpoint of the main axis. Moreover, although the example which uses the data table which matched angle ((theta), (phi)) with the combination of inclination direction and inclination angle was shown in the above-mentioned example, angle ((theta), (phi)) is used for one of an inclination direction and an inclination angle. It is also possible to narrow down the candidate reference images using a data table in which these ranges are associated. In this case, the number of candidate reference images is larger than when both the tilt direction and the tilt angle are used, but a reference distance image is selected from all candidate reference images stored in the reference image storage unit 15. The processing load can be greatly reduced than the case.

  By the way, when the area of the upper surface of the obstacle Ob is large and occupies most of the visual field of the distance image sensor 11 as shown in FIG. The reference distance image P2 having a high degree of coincidence with the upper surface may be selected. In this case, there is a possibility that the obstacle Ob cannot be found from the difference image between the distance image obtained by the distance image sensor 11 and the reference distance image selected by the reference distance image selection unit 13.

  Therefore, the determination unit 14 generates a difference image Dd between the reference distance image selected by the reference distance image selection unit 13 and the distance image obtained by the distance image sensor 11, and then the difference image Dd as shown in FIG. Is divided into a proximity region D2 close to the mobile body 2 and a remote region D3 far from the mobile body 2 (the dividing line is indicated by a one-dot chain line), and each included in the proximity region D2 and the remote region D3 The area of the coincidence region D1 is obtained. Here, if the obstacle Ob has a high degree of coincidence with the reference distance image, the area of the coincidence region (shaded portion) D1 is small in the proximity region D2, and the area of the coincidence region D1 is large in the far region D3. This is because the area occupied by the obstacle Ob in the distance image first increases in the distance region D3 as the moving body 2 gradually approaches the obstacle Ob.

  Based on the above characteristics, the determination unit 14 individually prepares thresholds th2 and th3 for the area of the coincidence region D1 for the near region D2 and the far region D3, respectively. When the area S2 of the matching region D1 is equal to or smaller than the threshold th2 (S2 ≦ th2) with respect to the proximity region D2, and the area S3 of the matching region D1 is equal to or larger than the threshold th3 (S3 ≧ th3) with respect to the far region D3, as shown in FIG. Therefore, it is determined that the matching area D1 in the difference image corresponds to the obstacle Ob. On the other hand, when the area S2 of the coincidence region D1 is larger than the threshold th2 (S2> th2) for the near region D2, and the area S3 of the coincidence region D1 is less than the threshold th3 (S3 <th3) for the far region D3, the coincidence region D1. It is determined that the area that is not corresponds to the obstacle Ob.

  The determination unit 14 performs coordinate conversion so that the region determined to be the obstacle Ob in the distance image obtained from the distance image sensor 11 is mapped to the three-dimensional space of the orthogonal coordinate system set for the mobile body 2. This data is delivered to the mobile body 2. The mobile body 2 performs an action corresponding to the obstacle Ob using the three-dimensional data delivered from the determination unit 14. The behavior of the mobile body 2 with respect to the obstacle Ob is defined as a rule in the mobile body 2. For example, the rule is set so that the movement direction is changed so as to avoid the obstacle.

  By the way, the mobile body 2 described above possesses map information of the travel range, and autonomously constructs itself so as to acquire its own position by appropriate means and to take action that matches the self position using the map information. It is configured as follows. Various means for acquiring the self-position have been proposed, for example, receiving radio waves from a plurality of base stations whose positions are known, and acquiring the self-position using information such as the arrival direction or the arrival time of the radio waves. A technique using radio navigation, a technique for acquiring the current self-position using a history of a travel distance and a travel direction from a known reference position, and the like have been proposed.

  On the other hand, in order for the mobile body 2 to travel autonomously, map information for collating the acquired self-location is required, and the map information includes information on an obstacle whose position is fixed (that is, does not move) in advance. It is desirable to register. As described above, the present embodiment has a function of detecting an obstacle and acquiring the position of the obstacle. Therefore, the position of the obstacle is specified by using this function and the function of acquiring the self-position. Map information can be generated.

  Below, the procedure which produces | generates map information is demonstrated easily. As shown in FIG. 8, the mobile body 2 generates map information based on the position detection device 21 for acquiring its own position, the output of the image processing device 10, and the output of the position detection device 21. It is assumed that a map creation unit 22 and a map storage unit 23 that stores the map created by the map creation unit 22 are provided. The map information is information related to the position and size of the object in the absolute coordinate system set regardless of the mobile body 2. Here, information such as a building whose position and size in the absolute coordinate system are known is previously stored in the map storage unit 23 as map information.

  As described above, when the determination unit 14 detects the obstacle Ob while the mobile body 2 is moving, the position of the obstacle Ob in the coordinate system set for the distance image sensor 11 is delivered to the map creation unit 22. Since the determination unit 14 determines the presence or absence of the obstacle Ob based on the difference image between the distance image and the reference distance image, the reference plane on which the obstacle Ob exists is recognized. That is, this reference plane can be regarded as a road surface (or a ground surface or a floor surface).

  The map creation unit 22 performs coordinate conversion in which the reference plane is regarded as the xy plane in the absolute coordinate system for the area where the obstacle Ob exists based on the distance information between the moving body 2 and the obstacle Ob obtained from the distance image. Do. At this time, the height information of the obstacle Ob from the reference plane and the size in the xy plane are also obtained. By such coordinate transformation, the relative position of the obstacle Ob with respect to the moving body 2 on the reference plane can be obtained.

  On the other hand, since the position of the mobile body 2 is recognized by the position detection device 21, the map creation unit 22 determines the relative position between the mobile body 2 and the obstacle Ob obtained as described above, and the position. The position of the obstacle Ob in the absolute coordinate system is obtained using the self position obtained by the detection device 21. The map information regarding the position and size of the obstacle Ob thus obtained is stored in the map storage unit 23.

  Here, the reference plane adopted by the determination unit 14 corresponds to the road surface (or the ground or floor surface), but the road surface (or the ground or floor surface) is inclined with respect to the xy plane in the absolute coordinate system. Therefore, there is a possibility that an error is included in the position coordinates of the obstacle Ob obtained as described above. However, since the map information stored in the map storage unit 23 is map information used when the mobile body 2 travels, it does not necessarily reflect the exact position in the absolute coordinate system. That is, it is only necessary to be able to recognize the position of the obstacle Ob when the mobile body 2 actually travels. Therefore, the position error of the obstacle Ob in the map storage unit 23 causes no problem in actual use.

It is a block diagram which shows embodiment of this invention. It is an external appearance perspective view same as the above. It is operation | movement explanatory drawing same as the above. It is operation | movement explanatory drawing same as the above. It is operation | movement explanatory drawing which shows the procedure which selects the candidate reference | standard image in the same as the above. It is operation | movement explanatory drawing same as the above. It is operation | movement explanatory drawing same as the above. It is a block diagram which shows other embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Obstacle detection apparatus 2 Mobile body main body 10 Image processing apparatus 11 Distance image sensor 12 Difference image generation part 13 Reference distance image selection part 14 Judgment part 15 Reference image memory | storage part 16 Candidate image narrowing down part

Claims (6)

  A distance image sensor that is mounted on the mobile body and includes a spatial image including a surface on which the mobile body is about to run diagonally downward and generates a distance image with pixel values as distance values, and the mobile body travels on a plane. A reference image storage unit storing a distance image of a plane within an angle range predicted to be generated by the distance image sensor as a candidate reference image at a predetermined step size, and a plurality of images stored in the reference image storage unit A reference distance image that selects, as a reference distance image, a candidate reference image in which the area of a matching area whose difference value is equal to or less than a predetermined threshold in the difference image from the distance image generated by the distance image sensor among the candidate reference images of A movement comprising: a selection unit; and a determination unit that determines that an area other than the matching area in the difference image with respect to the reference distance image selected by the reference distance image selection unit is an area where an obstacle exists. Obstacle detection device use.   Difference image generation having a function of obtaining a matching area in which a difference value is equal to or less than a predetermined threshold in a difference image between a default candidate reference image stored in the reference image storage unit and a distance image generated by the distance image sensor And a direction orthogonal to the principal axis of the coincidence area is used to estimate the inclination direction of the traveling surface of the moving body, and to extract candidate reference images that are within a predetermined range with respect to the inclination direction from the reference image storage unit The obstacle detection device for a moving body according to claim 1, further comprising an image narrowing-down unit, wherein the reference distance image selecting unit selects a reference distance image from the candidate reference images narrowed down by the candidate image narrowing-down unit. .   Difference image generation having a function of obtaining a matching area in which a difference value is equal to or less than a predetermined threshold in a difference image between a default candidate reference image stored in the reference image storage unit and a distance image generated by the distance image sensor And a width of the coincidence area in a direction orthogonal to the principal axis of the coincidence area, the inclination angle of the traveling surface of the moving body is estimated, and a candidate reference image that is within a predetermined range with respect to the inclination angle is determined as the reference image. 2. The mobile object according to claim 1, further comprising a candidate image narrowing-down unit that is extracted from the storage unit, wherein the reference distance image selecting unit selects a reference distance image from the candidate reference images narrowed down by the candidate image narrowing-down unit. Obstacle detection device.   Difference image generation having a function of obtaining a matching area in which a difference value is equal to or less than a predetermined threshold in a difference image between a default candidate reference image stored in the reference image storage unit and a distance image generated by the distance image sensor And the direction perpendicular to the principal axis of the coincidence area is used to estimate the inclination direction of the running surface of the mobile body and the width of the coincidence area in the direction perpendicular to the principal axis of the coincidence area is used to travel the mobile body. A candidate image narrowing unit that estimates a tilt angle of a surface and extracts a candidate reference image that is within a predetermined range with respect to a combination of a tilt direction and a tilt angle from the reference image storage unit, and the reference distance image selection unit is a candidate image The obstacle detection device for a moving body according to claim 1, wherein a reference distance image is selected from candidate reference images narrowed down by a narrowing-down unit.   2. The reference distance image selecting unit selects a reference distance image from candidate reference images within a range defined for a previously used reference distance image with respect to a distance image obtained as a time series. The obstacle detection apparatus for moving bodies described.   A distance image sensor that is mounted on the mobile body and includes a spatial image including a surface on which the mobile body is about to run diagonally downward and generates a distance image with pixel values as distance values, and the mobile body travels on a plane. A reference image storage unit storing a distance image of a plane within an angle range predicted to be generated by the distance image sensor as a candidate reference image at a predetermined step size, and a plurality of images stored in the reference image storage unit A reference distance image that selects, as a reference distance image, a candidate reference image in which the area of a matching area whose difference value is equal to or less than a predetermined threshold in the difference image from the distance image generated by the distance image sensor among the candidate reference images of The selection unit and the matching area are divided into a proximity area close to the mobile body and a remote area far from the mobile body, and the area occupied by the matching area in the proximity area is equal to or less than the first threshold and When the condition that the area occupied by the coincidence area is equal to or greater than the second threshold is satisfied, the coincidence area is determined as an area where an obstacle exists, and when the condition is not satisfied, the area selected by the reference distance image selection unit is selected. An obstacle detection apparatus for a moving body, comprising: a determination unit that determines an area other than a matching area in a difference image with respect to a reference distance image as an area where an obstacle exists.

Images