JP4568637B2 - Road recognition device - Google Patents

Description

  The present invention relates to a road recognition device, and more particularly to a road recognition device capable of extracting a pedestrian crossing from a captured image obtained by imaging the front of a vehicle with a stereo camera.

  In recent years, road shapes have become a prerequisite for improving safety and automatic control by applying image processing to images captured by in-vehicle stereo cameras and video cameras for the purpose of improving driving safety and automatic control of vehicles, etc. Development of a road recognition device that recognizes (see, for example, Patent Documents 1 to 3).

To recognize the shape of the road on which the vehicle is traveling, recognize the position of lanes such as the overtaking prohibition line marked on the road surface in front of the host vehicle and the lane marking that divides the roadside zone from the roadway. Is important. Therefore, in such a road recognition device, usually, focusing on the luminance of the pixel of the captured image, the pixel portion where the luminance changes greatly using the fact that the luminance of the pixel greatly differs between the roadway and the lane is extracted as an edge. It is configured to recognize the lane.
JP 2001-92970 A JP-A-5-347000 JP 2004-310522 A

  However, in the road recognition device based on the camera information as described above, there is a possibility that a road marking such as a pedestrian crossing having luminance information similar to the lane is erroneously recognized as a lane. In particular, as in the lane recognition device described in Patent Document 1, in the case of lane recognition, there is a problem as described below when it is a condition that the interval between the high luminance portions is equivalent to the width of the lane.

  That is, although the standard of the lane is set to 10 cm to 20 cm, there is a thick lane up to about 10 cm to 55 cm on an actual road, so if the threshold of the lane detection width is set to 10 cm to 20 cm Cannot recognize thick lanes. On the other hand, if the threshold value of the lane detection width is enlarged to about 10 cm to 55 cm in accordance with the actual environment, a pedestrian crossing whose thickness is about 50 cm is recognized as a lane. For this reason, it is difficult to set a threshold value for distinguishing between a pedestrian crossing and a lane by width.

  Moreover, if the detection range of the high-intensity part in the lane recognition is widened, a narrowly marked pedestrian crossing may be erroneously recognized as a roadway.

  On the other hand, the technology for reliably recognizing and detecting pedestrian crossings where there are many pedestrians is extremely important from the viewpoint of ensuring the safety of pedestrians and judging the risk of crossing pedestrians when viewed from the vehicle side. This is a field where technology and development are strongly desired.

  The present invention has been made in view of the above circumstances, and an object thereof is to provide a road recognition device capable of accurately recognizing the presence and position of a pedestrian crossing from an image taken in front of a vehicle.

In order to solve the above problem, the first invention provides:
In the road recognition device,
Imaging means for imaging a vehicle periphery including a road and outputting a pair of images;
Conversion means for converting the pair of images into a pair of digital images having a predetermined luminance gradation for each setting region;
Image processing means for calculating a distance in real space for each setting region of at least one digital image based on the pair of digital images;
Have a detection means for detecting a pedestrian crossing in the luminance and the upper one of the digital image based on the distance of each setting area for the one digital image,
The detection means recognizes a repetitive pattern of luminance for each of the setting areas with a front area that is a predetermined distance away from the vehicle in real space as a detection target,
By determining whether or not a point in the real space corresponding to the setting area where the repetitive pattern of luminance is recognized is within an allowable range near the predetermined distance, a point in the real space corresponding to the setting area is determined. It is determined whether it is on the road surface, and when the point of the real space corresponding to the setting area is on the road surface by the determination, the repeated pattern of the brightness is detected as a crosswalk .

According to the first invention, in the road recognition device, the imaging unit images the periphery of the vehicle including the road ahead of the vehicle and outputs a pair of images, and the conversion unit converts the pair of images to a predetermined luminance level for each pixel. A pair of digital images having a tonal brightness, image processing means calculates the distance in real space for each pixel of one digital image based on the pair of digital images, and the detection means for the one digital image A crosswalk is detected on one digital image based on the brightness and distance data for each pixel.
More specifically, the detection means recognizes a repetitive luminance pattern for each setting region with a front region that is a predetermined distance away from the host vehicle in real space as a detection target, and a setting region in which the repetitive luminance pattern is recognized. To determine whether the real space point corresponding to the setting area is on the road surface by determining whether or not the real space point corresponding to is within the allowable range near the predetermined distance, As a result of the determination, when a point in the real space corresponding to the setting area is on the road surface, a repeated luminance pattern is detected as a pedestrian crossing.

The second invention is the road recognition apparatus of the first aspect of the invention, said detecting means, a vehicle surroundings including road, detected a front region including the travel path and a travel path adjacent thereto on which the vehicle is traveling It is characterized by.

According to the second invention, the detection means of the road recognition device detects not only the traveling road on which the host vehicle is traveling, but also a front area on the traveling road adjacent thereto from the image captured by the imaging means. Search for and detect crosswalks.

According to a third aspect of the present invention, in the road recognition device according to the first or second aspect , the setting of the predetermined distance from the host vehicle is variable between the daytime and the nighttime in the front area to be detected. Features.

According to the third invention, the predetermined distance in the first or second invention is set to a different distance between daytime and nighttime, for example, an area 50m ahead in the daytime and an area 20-30m ahead in the nighttime. To do .

According to a fourth aspect of the present invention, in the road recognition device according to any one of the first to third aspects, the front area to be detected corresponds to an epipolar line having a width of one pixel on the one digital image. It is a region .

According to the fourth invention, as the front area in the first to third inventions, an area of an epipolar line having a width of one pixel is searched for in all pixel areas of one digital image serving as a reference of the pair of digital images. Then, a pedestrian crossing is detected by monitoring only a region away from the own vehicle in the real space by a predetermined distance corresponding to the epipolar line having a width of one pixel .

A fifth invention, in the road recognition apparatus of the first aspect of the invention, said detecting means, said detected as a repeat of the high luminance region over determination threshold repeating pattern predetermined brightness, between the high-brightness region When the interval is substantially equal to the width of the high-luminance region or approximately three times the width of the high-luminance region, the repeated pattern of the luminance is detected as a pedestrian crossing .

According to the fifth aspect of the invention, the detecting means detects the repetition pattern of the luminance as a repetition of the high luminance region equal to or higher than a predetermined determination threshold, and the interval between the high luminance regions is substantially equal to the width of the high luminance region or When it is approximately three times the width of the high-luminance area, a repeated pattern of luminance is detected as a pedestrian crossing.

  According to the first invention, the road recognition device generates luminance data for each pixel by converting the pair of images captured by the imaging unit into a pair of digital images having a luminance of a predetermined luminance gradation by the converting unit. To do. The image processing means calculates a distance in real space for each pixel of one of the digital images serving as a reference based on the pair of digital images, and generates distance data for each pixel. The detecting means detects a pedestrian crossing on the image based on the brightness and distance data for each pixel generated for the one digital image.

  Therefore, in correspondence with the standard width of the white line of the pedestrian crossing being about 45 to 50 cm and wider than the standard width of the lane 10 to 20 cm, the high-luminance pixel region appearing on the pixel, that is, the high-luminance region is discriminated by the width. Therefore, the white line and lane of the pedestrian crossing can be clearly distinguished and recognized, and the width of the white line can be accurately detected, so the presence of the pedestrian crossing can be accurately recognized without misrecognizing it as a lane. It becomes possible.

Further , the detection means of the road recognition device may detect that the high-luminance pixel area repeatedly appears on one of the reference digital images in the detection of the pedestrian crossing, and the real-space corresponding to the high-luminance pixel area. When it is recognized that the point is on the road surface, the repeated high brightness area is detected as a pedestrian crossing for the first time.
Therefore, high-brightness pixels reflecting guardrails and telephone poles that are higher than the road surface, pillars and bumpers of the preceding vehicle, and dust and dirt on the road surface that do not have repetitive patterns. It is possible to prevent a pixel from being erroneously determined as a white line of a pedestrian crossing.

In addition, the detection means of the road recognition device does not detect a pedestrian crossing by searching all pixel regions of one of the digital images serving as a reference of the pair of digital images, but is a predetermined distance from the vehicle in real space. Only the area that is far away is monitored to detect pedestrian crossings. In this way, the processing time can be remarkably shortened as compared with the case of searching for all the pixel areas of one digital image serving as a reference.
  In addition, pedestrian crossings approaching from a distance as the vehicle travels always pass through a region that is a predetermined distance away from the vehicle, so it is possible to reliably cross the pedestrian crossing by monitoring the region ahead of the vehicle by a predetermined distance. Can be detected.

According to the second invention, the detection means of the road recognition device detects not only the traveling road on which the host vehicle is traveling, but also a front area on the traveling road adjacent thereto from the image captured by the imaging means. For example, even if the preceding vehicle is on the pedestrian crossing and the pedestrian crossing cannot be detected on the traveling road on which the vehicle is traveling, A pedestrian crossing can be detected, and the effects of the first invention can be more accurately exhibited.

According to the third invention, by setting the predetermined distance in the first or second invention to a different distance between daytime and nighttime, for example, in the daytime, an area 50 m ahead is monitored from the viewpoint of early detection of a pedestrian crossing. It is possible to make the effects of the first or second invention appear more accurately by setting so as to monitor the area in front of 20-30m that is reliably illuminated by the headlight at night. At the same time, such information can be sent to various control units such as an automatic transmission (AT) control unit so that various control units can take appropriate measures, thereby ensuring the safety of crossing pedestrians more reliably. Is possible.

According to the fourth invention, as the front area in the first to third inventions, an area of an epipolar line having a width of one pixel is searched for in all the pixel areas of one digital image serving as a reference of the pair of digital images. Then, a pedestrian crossing is detected by monitoring only a region separated from the host vehicle by a predetermined distance corresponding to the epipolar line having a width of one pixel in the real space. Therefore, it takes time to detect compared to the case of searching the epipolar lines of the two or more pixels width is shortened, it becomes possible to clearly detect the position of the pedestrian crossing, the third from the first The effect of the invention is more accurately exhibited.

  Hereinafter, an embodiment of a road recognition device according to the present invention will be described with reference to the drawings.

  FIG. 1 is a block diagram of a road recognition device according to the present embodiment. The road recognition device 1 is mainly composed of an imaging means 2, a conversion means 3, an image processing means 6, and a detection means 9.

  The imaging means 2 images the periphery of the vehicle, and is configured to capture a landscape including a road ahead of the vehicle and output a pair of images. In the present embodiment, a stereo camera including a pair of main camera 2a and sub-camera 2b each incorporating a synchronized image sensor such as a CCD or CMOS sensor is used.

  The main camera 2a and the sub camera 2b are attached, for example, in the vicinity of the rearview mirror with a predetermined interval in the vehicle width direction, and each of the pair of stereo cameras closer to the driver will be described later. A main camera 2a that captures an image on which a distance is calculated for a pixel and a crosswalk is detected, and a sub-camera 2b that captures an image that is compared with a camera far from the driver to obtain the distance or the like. ing.

  A / D converters 3a and 3b as conversion means 3 are connected to the main camera 2a and the sub camera 2b, respectively, and the A / D converters 3a and 3b are output from the main camera 2a and the sub camera 2b. The pair of analog images is configured to convert, for each pixel, a reference image, which is a digital image having a predetermined luminance gradation such as a gray scale of 256 gradations, and a comparative image.

  An image correction unit 4 is connected to the A / D converters 3a and 3b, and the image correction unit 4 applies the main camera 2a and the comparison image to the reference image and the comparison image output from the A / D converters 3a and 3b. Image correction such as displacement due to an error in the mounting position of the sub camera 2b and correction of luminance is performed using affine transformation or the like.

  In the present embodiment, the reference image is image data having a luminance of 512 pixels in the horizontal direction and 200 pixels in the vertical direction as shown in FIG. 2, and the comparison image is 640 pixels in the horizontal direction and vertical. The image correction unit 4 is configured to output the image data having a luminance of 200 pixels as a direction. Each image data is stored in an image data memory 5 connected to the image correction unit 4. In addition, the dot part in FIG. 2 represents main shadows, such as a vehicle and an electric pole.

  An image processing unit 6 is connected to the image correction unit 4, and the image processing unit 6 mainly includes an image processor 7 and a distance data memory 8.

  The image processor 7 includes each pixel or a plurality of pixels of the reference image based on the digital data of the reference image and the comparison image output from the image correction unit 4 by the stereo matching processing and filtering processing output from the image correction unit 4. The parallax d is calculated in order to calculate the distance in the real space for each setting area including the blocks to be processed. The calculation of the parallax d is described in detail in Japanese Patent Application Laid-Open No. H5-114099 previously filed by the applicant of the present application.

  The image processor 7 calculates one parallax d for each pixel block of 4 × 4 pixels with respect to a reference image having 512 × 200 pixels. A luminance p1ij of 0 to 255 is assigned to each of the 16 pixels constituting one pixel block, and the luminance p1ij of the 16 pixels forms a luminance characteristic unique to the pixel block. The subscripts i and j of the luminance p1ij are the i and j coordinates of each pixel when the lower left corner of the image plane of the reference image or comparison image is the origin, the horizontal direction is the i coordinate axis, and the vertical direction is the j coordinate axis. To express.

  In the stereo matching process in the image processor 7, the reference image is divided into a maximum of 128 × 50 pixel blocks every 4 × 4 pixels as described above, and the comparison image is formed into an epipolar line having a width of 4 pixels extending in the horizontal direction. To divide. Then, one pixel block of the reference image is taken out and the city block distance CB obtained by the following equation (1) is minimized while shifting the pixel on the epipolar line of the corresponding comparison image in the horizontal direction, that is, in the i direction. The pixel block on the epipolar line, that is, the pixel block on the comparative image having luminance characteristics similar to the pixel block of the reference image is searched.

なお、ｐ２ijは比較画像上の座標（ｉ，ｊ）の画素の輝度を表す。

Note that p2ij represents the luminance of the pixel at the coordinates (i, j) on the comparison image.

  The image processor 7 calculates a shift amount between the pixel block on the comparison image specified in this way and the pixel block on the reference image, and assigns the shift amount to the pixel block on the reference image as a parallax d. It has become. Note that the parallax d is a relative amount of relative displacement in the horizontal direction with respect to the mapping position of the same object in the reference image and the comparison image derived from a certain distance from the main camera 2a and the sub camera 2b. The distance from the center position of the sub camera 2b to the object and the parallax d are uniquely associated based on the principle of triangulation.

  Further, for the purpose of improving the reliability of the parallax d, the image processor 7 performs a filtering process on the parallax d thus obtained, and outputs only the parallax d that has been validated. That is, for example, even when a pixel block of 4 × 4 pixels having only a feature image of a roadway is scanned on the epipolar line of the comparison image, all the portions where the roadway of the comparison image is imaged are highly correlated, Even if the corresponding pixel block is specified and the parallax d is calculated, the reliability of the parallax d is low. Therefore, such parallax d is invalidated in the filtering process, and 0 is output as the value of parallax d.

  Therefore, the distance data of each pixel of the reference image output from the image processor 7, that is, the parallax d for calculating the distance in the real space for each pixel block of the reference image is normally as shown in the pseudo image of FIG. In addition, the data has an effective value only for a so-called edge portion where the difference in luminance p1ij is large between pixels adjacent in the left-right direction of the reference image. In FIG. 3, pixel blocks having distance data are represented in black.

  The distance data of each pixel block of the reference image calculated by the image processor 7 is stored in the distance data memory 8 of the image processing means 6.

  In this embodiment, the detection means 9 is composed of a microcomputer in which a CPU, ROM, RAM, input / output interface, etc. (not shown) are connected to the bus. For example, the detection means 9 may be built in an existing EPU in the vehicle. Is possible. The detection means 9 detects a pedestrian crossing on the reference image based on the luminance and distance of each pixel with respect to the reference image.

  The detection means 9 also has a function of extracting a road lane from each data stored in the image data memory 5 and the distance data memory 8. The detection result of the pedestrian crossing and the lane is the detection means 9. Are transmitted to various control units such as an automatic transmission (AT) control unit, an engine control unit, an anti-lock brake system (ABS) control unit, a traction control system (TCS) control unit, a vehicle behavior control unit, etc. (not shown) Appropriate processing is performed on the basis of this in the unit and transmitted to display means such as an in-vehicle monitor or displayed, or transmitted to an alarm device to issue a necessary warning to the driver.

  The lane extraction in the detection means 9 can be performed using a known method. In the present embodiment, the detection means 9 is configured to have the function of the lane recognition device described in Patent Document 1. Yes.

  The detection means 9 stores image data composed of the luminance p1ij of each pixel of the reference image stored in the image data memory 5 and the distance data memory 8 and data of parallax d which is distance data for each pixel block of the reference image. Each is read out and a pedestrian crossing on the reference image is detected based on the data.

  In the present embodiment, the detection means 9 detects a pedestrian crossing searched for all pixel areas of the forward landscape imaged in the reference image, so that the processing time becomes long. Therefore, the front area separated from the vehicle in the real space by a predetermined distance. It is configured to search only for crosswalks. For example, the area ahead of the pedestrian crossing is set 50m ahead from the viewpoint of early detection of the pedestrian crossing in the daytime, and the area 20-30m ahead illuminated by the headlight is set at night. It is preferable if the setting can be switched as described above.

  Specifically, first, when the distance L from the own vehicle to the front area to be searched in the initial setting, for example, 50 m in the above example, is set, the detection means 9 first sets a reference image on the reference image corresponding to the front area. A j-coordinate jr of an epipolar line having a width of one pixel is calculated.

That is, as shown in FIG. 4, when the focal length of the main camera 2a is f and the mounting height of the main camera 2a is h, the shift between the image of the front region R and the vanishing point at the imaging position in the main camera 2a. Since x is expressed by x = h × f / L, on the reference image, the front region R has a pixel length of p,
xpixel = x / p = h × f / (L × p) (2)
Only represented as epipolar lines located below the vanishing point. Accordingly, by subtracting the xpixel from the j coordinate ja of the vanishing point determined by the attachment state of the main camera 2a, the j coordinate jr of the epipolar line on the reference image corresponding to the distance L to the front region R is calculated. Hereinafter, this epipolar line is referred to as an epipolar line jr.

  In the present embodiment, as described above, the front area to be detected is an area corresponding to an epipolar line having a width of 1 pixel on the reference image, but corresponds to an epipolar line having a width of 2 pixels or more. It is also possible to search for a forward area to be searched.

  In this embodiment, when the detection unit 9 reads the image data of the reference image from the image data memory 5, the detection unit 9 moves rightward from the leftmost i = 0 pixel on the epipolar line jr of the reference image as shown in FIG. It is configured to read the luminance p1ijr of all pixels. That is, in the present embodiment, not only the travel path on which the host vehicle is traveling, but also the travel paths adjacent to the travel path are targeted for detection of pedestrian crossings.

  The luminance p1ijr read in this way has a distribution as shown in FIG. 6, for example. In FIG. 6, the luminance p1ijr does not necessarily correspond to the luminance obtained by scanning the epipolar line jr shown in FIG.

  Conditions under which the detection means 9 determines that a pedestrian crossing exists on the epipolar line jr based on the luminance p1ijr are as follows.

Condition 1: A pixel whose luminance differential value is equal to or greater than the positive threshold Eth and whose luminance is equal to or greater than the start point luminance determination threshold Ath1 among the pixels on the epipolar line jr. Hereinafter, this pixel is referred to as a start pixel, and a point on the real space corresponding to the start pixel is referred to as a start point.
Condition 2: The starting point is on the road surface.
Condition 3: A pixel having a luminance differential value equal to or less than a negative threshold −Eth is present in pixels on the epipolar line jr. Hereinafter, this pixel is referred to as an end pixel, and a point on the real space corresponding to the end pixel is referred to as an end point.

Condition 4: There is a high luminance region in which the interval between the start point and the end point is within a predetermined range, and the luminance average value Aave from the start pixel to the end pixel is equal to or greater than the white line luminance determination threshold Ath2.
Condition 5: The interval between adjacent high-brightness areas is equal to the width of the high-brightness area or three times the width of the high-brightness area.
Condition 6: There are three or more high luminance areas.

  The limitation on the luminance differential value at the start pixel under Condition 1 is due to the fact that a strong edge strength appears in the luminance characteristics at the boundary between the asphalt of the roadway and the white line of the pedestrian crossing. Moreover, even if the luminance differential value condition is cleared, for example, the boundary between the pixel that reflects the shadow and the pixel that reflects the asphalt of the roadway, etc., as shown in FIG. In this case, the pixel is not determined as a start pixel.

  The detecting means 9 takes the difference between the luminance p1ijr of each pixel on the epipolar line jr of the reference image and the luminance p1i-1jr of the immediately preceding pixel to determine whether or not the pixel satisfies the condition 1. Yes.

  The condition 2 is intended to be excluded from the start pixel even when the pixel clears the condition 1, for example, when the pixel is a guardrail, a telephone pole, a pillar or a bumper of a preceding vehicle, or the like. . Since the guard rail, the preceding vehicle, and the like are at a position higher than the road surface, if it is on the epipolar line jr, that is, on the line connecting the front region R and the main camera 2a in FIG. The distance to etc. is smaller than the distance L to the front area R. Therefore, it is possible to determine whether or not the point in the real space corresponding to the pixel is on the road surface by examining the distance data of the pixel that clears the condition 1.

  The detection means 9 reads the distance data from the distance data memory 8 for the pixels satisfying the condition 1, calculates the distance in the real space, and whether or not the calculated distance is within the allowable range near the set distance L. For example, only a pixel within the allowable range such as a pixel A in FIG. 6 is determined as a start pixel. A pixel or the like that reflects the bumper portion of the preceding vehicle, such as the pixel T in FIG. 6, is excluded from the start pixel at this stage. Note that the allowable range is appropriately determined from the accuracy of the parallax d, the pixel length p, and the like.

  Condition 3 and Condition 4 find the end point that is the end opposite to the start point of the white line of the pedestrian crossing, and the interval between the start point and the end point and the brightness of the pixels between them are appropriate as the white line of the pedestrian crossing It is for judging whether or not. The standard of the width of the white line of the pedestrian crossing is about 45 to 50 cm, which is wider than the standard of the lane 10 to 20 cm. Therefore, for example, if the distance between the start point and the end point including an error or the like is within a range of 25 to 70 cm in real space, it can be determined as a white line of a pedestrian crossing.

  The detection unit 9 determines whether each pixel on the epipolar line jr satisfies the condition 3 in the same manner as the determination of the condition 1, and ends the pixel that satisfies the condition 3 such as the pixel B in FIG. The distance is calculated by reading the distance data for the end pixel from the distance data memory 8 from the distance data memory 8, and the distance in real space between the start point corresponding to the start pixel and the end point corresponding to the end pixel is the predetermined range. And whether or not the average luminance value Aave from the start pixel to the end pixel is equal to or greater than the white line luminance determination threshold Ath2.

  Here, as shown in FIG. 7, the distance between the start point and the end point in the real space is such that the start point P is projected onto the front area so as to correspond to the start pixel on the reference image, and similarly the end pixel. The end point Q is projected so as to correspond to the above, and the distance between the points P and Q is calculated.

  The threshold value Eth of the luminance differential value is appropriately determined. In this embodiment, the setting of the start point luminance determination threshold value Ath1 and the white line luminance determination threshold value Ath2 is calculated based on the road surface luminance Aroad that is the luminance of the road surface. It has come to be.

  As described above, in this embodiment, the detection means 9 is configured to have the function of the lane recognition device described in Patent Document 1, and in the lane recognition, on the epipolar line of the reference image. In order to detect the lane, the luminance histogram of the pixel corresponding to the portion outside the lane in the vicinity of the lane in the epipolar line and the four epipolar lines immediately below the epipolar line is calculated, and the luminance value having the maximum appearance frequency is calculated. The road surface brightness is Aroad.

  In the present embodiment, the road surface brightness Aroad may be calculated in the same manner, but there are cases where the road surface brightness Aroad cannot be calculated because the four epipolar lines immediately below the epipolar line jr also show pedestrian crossings. In the present embodiment, for example, the road surface brightness Aroad used for lane detection in the tenth epipolar line below the epipolar line jr is used.

  The starting point luminance determination threshold Ath1 and the white line luminance determination threshold Ath2 are calculated by adding constants α and β (where 0 <α <β) to the road surface luminance Aroad. Note that the values of α and β are appropriately determined.

  Conditions 5 and 6 are used to determine whether or not the repetitive luminance pattern for each pixel recognized as described above is repeatedly detected at a certain interval, such as a white line on a pedestrian crossing. is there. As shown in FIG. 8 (A), the pedestrian crossing is ideally a white line that is repeatedly marked with the same width as the white line, such as a car or the like as shown in FIG. 8 (B). Even if some of the tire marks are not recognized as white lines, the distance between the detected white lines should be an odd multiple of the width of the white line.

  Therefore, in this embodiment, the detection means 9 calculates the distance between the start point and the end point of the high brightness area detected up to the condition 4 from the respective distance data read from the distance data memory 8, and The interval of the luminance area is equal to the distance or three times the width of the high luminance area, and when such a high luminance area is repeated three times or more, they are configured to be detected as a pedestrian crossing. Yes. As shown in FIG. 8C, even if the high luminance region, that is, the white line portion has the same width, if the interval does not satisfy the above condition, it is not regarded as a pedestrian crossing.

  It should be noted that an allowable range is provided for the above-described criterion of whether the interval between the white lines is equal to the width or tripled, and it is determined that the white line interval is equal or tripled within the range. It is like that. This range is appropriately determined.

  Next, the operation of the road recognition device according to the present embodiment will be described.

  FIG. 9 is a flowchart showing a basic routine of processing of the road recognition apparatus according to the present embodiment. When the road recognition device 1 is activated, first, distance data and image data are generated by processing in the image pickup means 2, the conversion means 3, the image correction unit 4, and the image processing means 6 (step S1), and the image data memory 5 and the distance data memory 8.

  Then, the detecting means 9 detects a pedestrian crossing based on the distance data and image data (step S2), and the detected data such as the start pixel and the end pixel are controlled by the automatic transmission (AT) control unit and the like. (Step S3) and utilized for processing in each control unit. Moreover, the road recognition apparatus 1 performs the above process repeatedly.

  The generation of the distance data and the image data is as described in the configuration of the image processing means 6 and the like described above, and is described in detail in the Patent Document 1 and Japanese Patent Application Laid-Open No. 5-11409. Hereinafter, detection of the pedestrian crossing (step S2) in the detection means 9 will be described.

  FIG. 10 is a flowchart showing a processing routine in the detection means of the road recognition device according to the present embodiment. The detection means 9 first sets the number n of high-luminance areas corresponding to the white lines of the pedestrian crossing to 0 (step S20), and then the pixels on the epipolar line jr read from the image data memory 5 satisfy the condition 1. It is determined whether or not it is satisfied (step S21).

  Immediately after the search is started, there are many cases where the pixel does not satisfy the condition 1 (step S21: NO). In this case, it is determined whether a start pixel candidate to be described later is registered (step S22). Since no start pixel candidate is registered in the storage area (step S22: NO), it is determined whether or not all pixels on the epipolar line jr have been searched (step S23: NO), and the search target is the pixel. The search is repeated while moving to the pixel on the right side of the epipolar line jr.

  When a pixel satisfying condition 1 is found (step S21: YES), it is subsequently determined whether or not the pixel satisfies condition 2 (step S24). If the condition 1 is satisfied but the condition 2 is not satisfied (step S24: NO), such as the above-described pixel showing the guardrail, the preceding vehicle, etc., the search is continued without being regarded as a start pixel candidate.

  If it is determined that a point in the real space corresponding to a pixel satisfying the condition 1 is on the road surface (step S24: YES), the pixel is a pixel that can be a start pixel, and thus is registered as a start pixel candidate. (Step S25). Specifically, the coordinates of the start pixel candidate on the reference image are stored. At this time, if there is a registered start pixel candidate, the existing start pixel candidate is deleted, and a newly found start pixel candidate is registered. Thus, if a new start pixel candidate is found before the end pixel is found, the existing start pixel candidate is deleted.

  In the search for pixels after the pixel registered as the start pixel candidate, since the start pixel candidate is registered in the storage area (step S22: YES), whether or not the condition 3 is satisfied, that is, is the end pixel A determination of whether or not is made (step S26). Then, pixels that are determined not to be the end pixel (step S26: NO) continue, and when the number of pixels from the start pixel candidate exceeds a predetermined pixel range (step S28: YES), the start pixel candidate in the storage area is deleted. (Step S28).

  Here, the reason for setting the predetermined pixel range is that a start pixel candidate in which no end pixel is found even if a pixel range corresponding to a certain distance is searched is a pixel corresponding to the start point of the white line of the pedestrian crossing. It is not appropriate to screen such starting pixel candidates. In addition, the predetermined pixel range is set, for example, to a pixel corresponding to a point 1 m away from a point corresponding to the start pixel candidate in real space.

  After the start pixel candidate is registered, if a pixel satisfying condition 3 is found (step S26: YES), whether or not the pixel satisfies condition 4 is satisfied, that is, it corresponds to the end point of the white line of the pedestrian crossing. It is determined whether or not the point is suitable (step S29). The interval between the start point and the end point is not within a predetermined range, or the luminance average value Aave from the start pixel to the end pixel is the white line luminance determination threshold Ath2. Otherwise (step S29: NO), the start pixel candidate is deleted because it is not appropriate as the start point of the white line of the pedestrian crossing (step S28).

  If it is determined that the pixel satisfies the condition 4 (step S29: YES), it is determined whether or not there are already registered start and end pixels (step S30), and if not (step S30: NO) this time The start pixel candidate found in the search is registered in the storage area together with the end pixel as a start pixel (step S31). Specifically, the coordinates of the start pixel and the end pixel on the reference image are stored. Then, the detection number n is set to 1 (step S32), the start pixel candidate registered in the storage area is deleted (step S33), and all the pixels on the epipolar line jr are not searched (step S23). : NO), the search for the next pixel is continued.

  If there are already registered start pixels and end pixels (step S30: YES), whether or not the condition 5 is satisfied, that is, whether or not the interval between adjacent high-intensity areas is equal to the width of the high-intensity areas or high It is determined whether or not it is three times the width of the luminance area (step S34). If the condition 4 is satisfied (step S34: YES), the storage area together with the end pixel is set to the start pixel candidate detected this time as the start pixel. (Step S35), the detection number n is increased by 1 (step S36), and the start pixel candidate registered in the storage area is deleted (step S33).

  When the condition 5 is not satisfied as in the white line group shown in FIG. 8C (step S34: NO), all the registered start pixels and end pixels are deleted and detected this time. The start pixel candidate is registered in the storage area together with the end pixel as a start pixel (step S37). Then, the detection number n is set to 1 (step S38), and the start pixel candidate registered in the storage area is deleted (step S33).

  In this embodiment, when the pixel does not satisfy the condition 5 as described above, all the start pixels and end pixels that have already been registered are deleted and the number of detections is reset. When there are three or more sets of start pixels and end pixels, it is possible not to erase from the storage area because there is a possibility of a pedestrian crossing. Further, when the pixel does not satisfy the condition 5, the start pixel and the end pixel that are already registered may be left and the start pixel and the end pixel detected this time may not be registered.

  When the search is completed for all the pixels on the epipolar line jr in this way (step S23: YES), whether or not the condition 6 is satisfied, that is, there are three or more high-luminance regions that satisfy the conditions 1 to 5 above. Then, it is determined whether or not there is a repeating pattern of luminance for each pixel (step S39), and if there are three or more (step S39: YES), it is registered in the storage area as a pedestrian crossing is detected (step S40). A set of coordinates on the reference image of the start and end pixels is output.

  If the detected number n of the high brightness area is less than 3 (step S39), it is assumed that no pedestrian crossing is detected (step S41). Even if the coordinates on the reference image of the start pixel and the end pixel are registered in the storage area. They are not output.

  As described above, according to the road recognition device 1 according to the present embodiment, paying attention to the fact that the standard width of the white line of the pedestrian crossing is about 45 to 50 cm and wider than the standard width of the lane 10 to 20 cm, Therefore, the existence of a pedestrian crossing can be accurately recognized without misrecognizing it as a lane.

  In addition, it is determined whether the high-intensity area has a repeated pattern based on the luminance of the pixel on the reference image, and whether the width of the high-intensity area that may correspond to the white line of the pedestrian crossing is within the standard width Since the determination of whether or not is based on the distance on the real section obtained by image processing, the position and distance of the point in the real space corresponding to the pixel, the repetition pattern can be accurately calculated, It is possible to detect a pedestrian crossing.

  At that time, it is possible to further improve the detection accuracy of the pedestrian crossing by making a determination while confirming whether or not the point corresponding to the pixel of the high brightness area is on the road surface based on the distance data. Furthermore, such accurate pedestrian crossing information is transmitted to each control unit such as an automatic transmission (AT) control unit, so that each control unit can perform appropriate processing based on the information. .

  In the present embodiment, a case has been described where one epipolar line jr on the reference image or one epipolar line that is variable between daytime and nighttime is searched for. For example, various distances to the front region to be searched can be set according to the use in each control unit, and the search can be made on a plurality of epipolar lines.

  In the present embodiment, the luminance of all the pixels is read from the leftmost pixel to the right direction on the epipolar line jr. However, in addition to this, for example, the left and right directions from the center position of the epipolar line jr in the i direction. It is also possible to configure each to scan toward the.

  Further, in the present embodiment, while the luminance of the pixels on the epipolar line jr is read out, the start pixel and the end pixel are registered and deleted at the same time, but the process starts until the search on the epipolar line jr ends. The pixel and end pixel may not be erased, and after the search is completed, the set of the start pixel and the end pixel based on the condition 5 may be selected.

  In FIG. 4, the method of calculating the epipolar line jr on the reference image corresponding to the distance L to the front area R of the host vehicle is described on the assumption that the road surface is a horizontal plane. Sometimes it is.

As described above, in the present embodiment, the detection means 9 is configured to have the function of the lane recognition device described in Patent Document 1, and in such a case, in the lane recognition, the main camera 2a and the sub-camera 2b, where the point on the road surface just below the center is the origin, the vehicle width direction is the X axis, the vehicle height direction is the Y axis, and the forward direction is the Z axis. Lane recognition is performed using a lane model related to the road height represented by the formula (refer to Patent Document 1 for a calculation method of c _L , c _R , d _L and d _R ).
Left lane Y = c _L · Z + d _L (3)
Right lane Y = c _R · Z + d _R (4)

When the road is inclined, when the epipolar line jr on the reference image is determined, first, c = (c) using the values of c _L , c _R , d _L and d _R calculated in the lane recognition. _When calculating the average values c and d by _L + c _R ) / 2 and d = (d _L + d _R ) / 2, respectively, then obtaining the number of pixels xpixel from the vanishing point on the reference image to the epipolar line jr Instead of using the equation (2),
xpixel = x / p = (h− (c · L + d)) × f / (L × p) (5)
By calculating xpixel using, it is possible to accurately calculate the epipolar line jr on the reference image corresponding to the distance L to the front area of the host vehicle.

  As described above, f represents the focal length of the main camera 2a, h represents the mounting height of the main camera 2a, and p represents the pixel length.

It is a block diagram of a road recognition device concerning this embodiment. It is a figure explaining the reference image converted into digital data. It is a figure which shows the pseudo image which arranged the distance data so that it might respond | correspond to a reference | standard image. It is a figure explaining a response | compatibility with the front area | region where the distance was set, and the epipolar line on a reference | standard image. It is a figure explaining the relationship between a reference image and an epipolar line. It is a figure which illustrates distribution of the brightness | luminance obtained by scanning on an epipolar line. It is a figure explaining the calculation method of the distance of the point on an epipolar line. It is a figure explaining the various repetitive patterns from which a white line is detected. It is a flowchart which shows the basic routine of the process of the road recognition apparatus which concerns on this embodiment. It is a flowchart which shows the processing routine in a detection means.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Road recognition apparatus 2 Imaging means 3 Conversion means 6 Image processing means 9 Detection means jr Epipolar line pnij Luminance

Claims (5)

Imaging means for imaging a vehicle periphery including a road and outputting a pair of images;
Conversion means for converting the pair of images into a pair of digital images having a predetermined luminance gradation for each setting region;
Image processing means for calculating a distance in real space for each setting region of at least one digital image based on the pair of digital images;
Have a detection means for detecting a pedestrian crossing in the luminance and the upper one of the digital image based on the distance of each setting area for the one digital image,
The detection means recognizes a repetitive pattern of luminance for each of the setting areas, with a front area that is a predetermined distance away from the vehicle in real space as a detection target,
By determining whether or not a point in the real space corresponding to the setting area where the repetitive pattern of luminance is recognized is within an allowable range near the predetermined distance, a point in the real space corresponding to the setting area is determined. A road recognition characterized by determining whether or not the road is on a road surface, and detecting the repetitive pattern of the luminance as a pedestrian crossing when a point in the real space corresponding to the set area is on the road surface apparatus. 2. The road recognition device according to claim 1, wherein the detection means detects, as a vehicle periphery including a road, a traveling area where the host vehicle is traveling and a front area including a traveling path adjacent thereto. . Before SL front region, a road recognition apparatus according to claim 1 or 2, characterized in that there is a variable setting of a predetermined distance from the vehicle in the daytime and nighttime. Before SL front region, a road recognition apparatus according to any one of claims 1 to 3, characterized in that the area corresponding to the epipolar line of one pixel width on the one digital image. The detection means detects the repetition pattern of the luminance as a repetition of a high luminance region equal to or greater than a predetermined determination threshold, and an interval between the high luminance regions is substantially equal to a width of the high luminance region or the high luminance region for a substantially three times the width, road recognition apparatus according to claim 1, characterized in that detecting the repetitive pattern of the luminance as a crosswalk.

Images