JP4793324B2 - Vehicle monitoring apparatus and vehicle monitoring method - Google Patents

Description

  The present invention relates to a vehicle monitoring apparatus for detecting a vehicle on a road and a program therefor.

  A camera is installed on a pillar installed on the road, and a video is taken of the road and the vehicle with the camera, and the video is processed to detect and track the vehicle and measure the number of passing vehicles and speed for each lane. Traffic volume measuring devices are known. The vehicle detected in the camera screen is the camera parameters of the camera, such as the camera height, pitch angle (decline angle), roll angle (rotation angle about the camera lens optical axis), yaw angle (road direction and camera). ), The focal length (related to the zoom ratio of the camera), and the like are converted into an actual coordinate system on the road surface, thereby determining the traffic lane and measuring the passing speed. For example, in the first embodiment of the object recognition apparatus described in Patent Document 1, an image from an overhead camera is projected onto a height of zero, that is, on a road surface based on the camera parameters of the camera. After extracting vehicle features (mainly horizontal edges) from the projected image, these horizontal edges are grouped by lane or according to a predetermined distance and applied to the vehicle model for each group to identify the position of the vehicle An object recognition apparatus is disclosed.

  FIG. 5 shows a pinhole camera model of a road camera. VP is the camera installation position as well as the camera viewpoint, and the distance F between the VP and the screen is the focal length. In reality, the height of the camera is several meters from the road, and the focal length F is several tens of millimeters. Usually, since the lens is focused, the center of the lens is placed at the viewpoint position, the screen is placed behind the lens, and a light receiving element such as a CCD is placed here to convert the image into an electrical signal. However, in FIG. 5, a screen is virtually placed in front of the viewpoint so that the perspective transformation of the pinhole camera model can be easily explained. For example, if the camera center is directed to a point gCNT on the road surface, the viewpoint VP and the screen centers cCNT and gCNT are in a straight line on the camera optical axis. Here, the pitch angle refers to an angle formed by a line segment connecting VP and gCNT and a line segment connecting O and gCNT, which is a position on the road of the column where the camera is installed. If the origin of the world coordinate system with the origin on the road is placed on O, the X-axis in the direction of crossing the road, the Y-axis in the direction of the road, and the Z-axis in the upward direction of the column, the line segment connecting the origin O and gCNT is the Y-axis. The angle formed is called the yaw angle. Although not shown here, the roll angle is the rotation angle of the screen about the camera optical axis (half line connecting VP and cCNT).

  FIG. 6 is an example of a road bird's-eye view image using a camera installed on a support column standing on the roadside. This is a case where a large vehicle exists in the first lane and a small vehicle exists in the second lane. Since the yaw angle is not zero, the larger vehicles with higher heights protrude more into adjacent lanes.

Japanese Patent No. 3516118 JP 2001-357402 A Japanese Patent No. 3541774 Japanese Patent No. 3435623 Seiji Iguchi: “Three-dimensional image measurement”: Shoshodo: pp91-99

  In Patent Document 1, since the height of an object cannot be measured for observation with one camera, it is assumed that each part of the object is on the road surface, and the vehicle is horizontal after being perspective-transformed on the road surface. An apparatus for detecting an edge is disclosed. FIG. 7A is a diagram obtained by perspective-transforming on the road surface according to the description in Patent Document 1, and FIG. 7B is a diagram in which a horizontal edge is further extracted. However, the lane markings and the vehicle contours indicated by broken lines in FIG. 7B are provided for convenience of explanation, and cannot be extracted by the horizontal edge detection process. In Patent Literature 1, the horizontal edge is classified for each vehicle and applied to various edge pattern models provided in advance to specify the position of the vehicle. However, the horizontal edges are grouped on a lane-by-lane basis or on the basis of the distance after perspective transformation. In particular, at the horizontal edge of a large vehicle in a scene with a yaw angle as in the example of FIG. There was a high possibility of mistakes, and there was a problem that the number of vehicles and their positions could not be measured appropriately.

  An object of this invention is to provide the vehicle monitoring apparatus which improved the measurement precision of the number of vehicles and the presence position using one camera.

  One of the features of the present invention is that an edge detection unit that detects a vehicle edge from an image captured by a camera, a projection unit that projects the vehicle edge detected by the edge detection unit on an arbitrary plane, and a projection unit that projects And an edge position estimating means for estimating a projection position where the edge length is equal to or less than the threshold when the edge length is larger than a predetermined threshold.

  Alternatively, vehicle edge detection means for detecting the edge of the vehicle from an image photographed by the camera, projection conversion means for projecting and transforming the image photographed by the camera on the road surface according to camera parameters, and projecting on the road surface by the projection conversion means A first estimation unit that estimates the position of the edge on the road surface as the vehicle existing position when the length of the vehicle edge of the captured image is smaller than a predetermined threshold value, and the projection conversion unit projects the image on the road surface. When the length of the vehicle edge in the image is larger than a predetermined threshold value, the position where the edge is projected is pulled back in the direction of the line of sight of the camera, and the projection position where the edge is equal to or less than the threshold value is The above-mentioned subject can be achieved by providing the 2nd estimating means which presumes the existence position of.

  Alternatively, when an edge of a vehicle is detected from an image captured by a camera, the detected edge of the vehicle is projected on a plane parallel to a road surface, and the length of the projected edge is larger than a predetermined threshold, The above problem can be solved by a vehicle monitoring method in which a projection position in which the length of the edge is the threshold value is estimated as the presence position of the edge.

  With the above configuration, it is possible to improve the accuracy of detecting the vehicle location, and there is an effect that it is possible to provide a highly accurate measuring device when measuring traffic indicators such as traffic volume.

  The best mode for carrying out the vehicle monitoring apparatus of the present invention will be described below with reference to the drawings.

  FIG. 1 shows an embodiment of a vehicle monitoring apparatus according to the present invention, in which an image input means 11, a vehicle rear narrowing means 12, a horizontal edge detecting means 13, a vehicle width converting means 14, a vehicle position estimating means 15, and a vehicle A tracking unit 16, a traffic index calculation unit 17, and a camera parameter holding unit 18 are included. The image input means 11 can use an ITV camera, and inputs it to the vehicle rear narrowing means 12 and the vehicle tracking means 16 at a constant frame rate of about 60 to 15 frames per second. The vehicle rear narrowing means 12 narrows down the rear region of the vehicle based on the presence or absence of a movement vector, a change in similarity with a road surface image, or a change in edge luminance from a plurality of frames of images input in time series. The horizontal edge detection means 13 detects the horizontal edge in the area narrowed down by the vehicle rear narrowing means 12. Next, the vehicle width conversion means 14 measures the length of the horizontal edge when projected onto the road surface by the perspective conversion method used in the pinhole camera model (detailed in Non-Patent Document 1). The vehicle position estimation means 15 determines the vehicle width threshold value in the range of the upper limit of 2.5 meters to 3.0 meters as the vehicle width of a large vehicle, and the length when projected on the road surface of the horizontal edge. Is equal to or less than the vehicle width threshold value, the projection position of the edge on the road surface is estimated as the rear position of the vehicle, and the length when projected on the road surface of the horizontal edge is the vehicle width threshold value. If it is larger than that, the horizontal edge is virtually pulled back in the direction approaching the camera along the line of sight of the camera of the image input means 11, and the vehicle rear position is directly below the road surface at a position that is equal to or less than the vehicle width threshold value. And a small image area is cut out based on the estimated position of the rear portion of the vehicle. Here, the vehicle width upper limit of the large vehicle is when the present invention passes through a dedicated gate such as a toll road, and the vehicle width of the passing vehicle is measured and known when passing through the dedicated gate. The vehicle width measured for a vehicle classified as a large vehicle can also be used. Also, a value that is 10% to 50% higher than the average value of a plurality of vehicles can be used. For example, as disclosed in Patent Document 4, the vehicle tracking unit 16 regards the small image area as a luminance pattern at the rear of the vehicle and uses it as a first template, and tracks the vehicle by pattern matching. The traffic index calculation means 17 determines the traffic lane from the position of the vehicle tracked by the vehicle tracking means 16 and counts the number of passing vehicles for each lane. Further, the speed of the vehicle is measured from the amount of movement of the vehicle tracked by the vehicle tracking means 16 for each frame. Since the position and amount of movement of the vehicle measured by the vehicle tracking means 16 are only on the screen of the camera of the image input means 11, it is necessary to project the position of the vehicle on the road surface by the perspective transformation method. The camera parameter holding unit 18 holds the camera parameters of the camera of the image input unit 11 and is referred to when the vehicle width conversion unit 14, the vehicle position estimation unit 15, and the traffic index calculation unit 17 perform perspective conversion.

  FIG. 2 is a diagram showing an example of the vehicle rear portion detection means. In FIG. 2, the vehicle rear narrowing means 12 includes a reference pattern holding means 21 that holds a reference brightness pattern of a road surface of a collation area provided at a position where a vehicle enters in the screen of the input image, an input image, and the reference A similarity buffer memory that is accessed based on a pattern matching means 22 for pattern matching using a correlation coefficient as a similarity, and a memory address management means 24 for performing cyclic addressing to hold a pattern matching similarity for a predetermined frame. 23, and vehicle rear candidate cutout means 25 that outputs the vehicle rear candidate image from the input image by measuring the passage timing of the vehicle rear from the similarity time series change held in the similarity buffer memory 23.

  FIG. 3 shows an example of the processing screen of the vehicle rear narrowing means 12, and the collation area 31, the collation area 32, and the collation area 33 are set at positions where the vehicle enters for each lane. In each collation area, a correlation calculation is performed for the reference pattern of the road surface held and the corresponding area of the input image, and the similarity is referred to. When there is no vehicle and the road surface appears on the image, the similarity is high, and when the road surface is shielded while the vehicle is passing, the similarity is low in principle. In the similarity time series, the timing at which the low-similarity frame continues for a while and the similarity becomes high is immediately after the vehicle passes through the matching region. At this timing, the region 34, the region 35, and the region 36 There is a high possibility that the part having the longest horizontal edge is the position where the rear part of the vehicle is in contact with the road surface. Therefore, when the vehicle passage timing is detected in the collation region 31, the collation region 32, and the collation region 33, the region 34, the region 35, and the region 36 are output, respectively. The above is an example of the operation of the vehicle rear narrowing means 12. In the verification area, the presence / absence of a movement vector may be used as the vehicle passage timing as disclosed in Patent Document 2, or the vehicle passage timing may be measured by a change in horizontal edge strength as disclosed in Patent Document 3. .

  FIG. 4 is a diagram showing an example of an image area to be processed by the horizontal edge detecting means 13 and detected edges. FIG. 4A shows a case where the processing area 34 is inputted at the correct timing as the rear part of the vehicle, and the detected edge here is a horizontal edge 41 shown in FIG. 4B. This is a position where the vehicle is in contact with the road surface, and is an edge suitable for specifying the vehicle position. On the other hand, FIG. 4C shows a case where the processing area 34 is inputted at an incorrect timing as the rear part of the vehicle, and the detected edges here are the horizontal edge 42 and the horizontal edge 43 shown in FIG. Such a timing error occurs when the vehicle brightness is close to the road surface brightness. If an edge having a large distance from the road surface, such as the horizontal edge 42 and the horizontal edge 43, is estimated as the position of the vehicle on the road surface, the presence lane of the vehicle is erroneous, and a large error is caused in speed measurement. In order to solve this problem, the processing of the vehicle width conversion means 14 and the vehicle position estimation means 15 in the next stage is required. As a filter for detecting a horizontal edge, there is a Sobel filter as shown in FIG. FIG. 13A shows the load index of the filter processing with the window size of 3 rows and 3 columns, and FIG. 13B shows the corresponding load. The Sobel filter is calculated by Equation 1.

  Here, In (x, y) is an image to be filtered, and indicates the luminance of a certain pixel in the area narrowed down by the vehicle rear narrowing means 12, where x and y are the screen of the pixel. Represents the coordinates above. Out (x, y) is the luminance of the corresponding pixel in the image after filtering. By performing the processing of Formula 1 with the load shown in FIG. 13B, the luminance difference in the vertical direction for each pixel in the filter target image is calculated. Therefore, the value of the horizontal edge portion having a large vertical luminance difference. Becomes larger.

  Here, the perspective transformation of the embodiment will be briefly described. The perspective transformation is a vehicle three-dimensional entity positioned in a world coordinate system in which the origin O is placed on the road side, the X and Y axes are defined on the road surface, and the Z axis is defined in the vertical direction of the road surface in the camera pinhole model shown in FIG. And a conversion that defines the correspondence of a two-dimensional image of a vehicle that is positioned in a screen coordinate system that is a two-axis orthogonal coordinate system (not shown) on the screen with the origin at the screen center cCNT. The vehicle detection by image processing originally only measures the position on the screen, so this is converted to the actual position in the world coordinate system by perspective transformation to measure the vehicle position and speed. FIG. 8 shows the relationship between the world coordinate system and the screen coordinate system. FIG. 8A is a diagram showing the conversion relationship between the world coordinate system and the screen coordinate system, and there is a camera coordinate system between the world coordinate system and the screen coordinate system. The camera coordinate system in FIG. 5 is a three-dimensional coordinate system in which one axis is provided on a line segment connecting VP and gCNT with VP as the origin, and the other two orthogonal axes are provided on a plane parallel to the screen. System (not shown). In perspective transformation, the world coordinate system and the screen coordinate system are related via this camera coordinate system. The transformation between the world coordinate system and the camera coordinate system is achieved by translational transformation and rotational transformation of the coordinate system. The camera coordinate system and the screen coordinate system are projected on the point where the line segment connecting the vehicle portion and the VP in the three-dimensional space (including the road surface) represented by the camera coordinate system intersects the screen plane. It can be obtained by proportional calculation using the distance F. When performing the above calculation, the camera parameters held by the camera parameter holding means 18 are used. The camera parameters are the pitch angle, the yaw angle, and the roll angle described in the camera pinhole model, the focal length F, and the world coordinates where VP is located, that is, (0, 0, h). Here, h is the height of the camera. Based on the above relationship, coordinates are converted by back projection in FIG. 8B and forward projection in FIG. Physically, since light reaches the screen from the object through the lens, it is opposite to the traveling direction of the light, but is defined as follows. By the back projection of FIG. 8B, the position of the vehicle portion in the three-dimensional space is converted as the position of the vehicle portion in the image. On the other hand, since the forward projection in FIG. 8C is a conversion from the two-dimensional coordinate system to the three-dimensional coordinate system, the position of the vehicle portion in the image is included in information such as the height or width of the vehicle portion. It is converted as the position of the part of the vehicle in the three-dimensional space.

  FIG. 9 is a diagram showing conversion of the horizontal edge 41, the horizontal edge 42, the horizontal edge 43, and the like into the world coordinate system by the vehicle width conversion means 14. The conversion is based on the forward projection in FIG. Since the height of the edge is unknown, the height is projected, that is, projected onto the road surface. In this state, FIG. 9A shows the road as seen from the side, and FIG. 9B shows the road as seen from directly above. In FIG. 9A, the actual position on the road at the rear of the vehicle is D. Therefore, when the horizontal edge generated at the position of C is converted into world coordinates based on an image (not shown) projected on the screen, the position of the world coordinates is in the vicinity of D, and the vehicle width measurement value based on this is It will be close to the actual vehicle width. On the other hand, when the horizontal edge generated at the position B is converted into world coordinates based on the image a projected on the screen, the position of the world coordinates is A. In FIG. 9 (b), it can be seen that A is displaced to the adjacent lane rather than the actual position of the rear portion of the vehicle, and the size is also enlarged. The same thing happens when the horizontal edge generated on the roof in the case of a small car is converted to the road surface, but the amount of displacement and expansion is small because of the low height.

  FIG. 10 is a diagram showing a processing flow of the vehicle position estimating means 15. As described above, when the detected horizontal edge is converted to world coordinates on the road surface, displacement and enlargement of the position occur, but small vehicles can be ignored, and large vehicles cannot be ignored. In view of this, the vehicle position estimating means 15 is a process for correcting the position displacement particularly for a large vehicle. In FIG. 10, the length of the horizontal edge is measured based on the world coordinates of the horizontal edge and compared with a predetermined value (st101). The predetermined value is about 2.5 meters, 3.0 meters, or 3.5 meters, and is an upper limit value of a general large vehicle. In general, when a vehicle with a width greater than this is not passing, the vehicle width greater than the predetermined value is measured. Judged to have been converted to. Therefore, if it is less than the predetermined value, it is determined as the world coordinate of the vehicle position as it is (st102). If it is larger than the predetermined value, a vehicle position pull-back estimation process is performed so that the length of the horizontal edge is equal to or smaller than the predetermined value (st103). Details of the vehicle position pullback estimation process will be described later. Thereafter, the vehicle position at this time is determined as the world coordinates of the vehicle position (st102). Next, the vehicle position is converted into a screen coordinate system by backprojecting the world coordinates of the vehicle in FIG. 8B (st103). Since the screen coordinate system of the vehicle position at st103 is originally the same as the output coordinates of the horizontal edge detecting means 13 on the route going straight from st101 to st102, this coordinate may be used as the processing result of st103.

  FIG. 11 is a conceptual diagram of the vehicle position pullback estimation process. FIG. 11A is a view of the road viewed from the side, and FIG. 11B is a view of the road viewed from directly above. The world coordinate Y-axis is determined in the road direction on the road surface, the X-axis is determined in the road crossing direction on the road surface, and the Z-axis is determined vertically above the road surface. The camera viewpoint VP is determined on the Z axis. In FIG. 11B, the edge α1α2 is obtained by projecting the edge onto the road surface by forward projection, and the world coordinates at both ends thereof are α1 (X1, Y1, 0) and α2 (X2, Y2, 0), respectively. The length is assumed to be R. The edge β1β2 has a length r and represents the position of the vehicle that is estimated to be pulled back so as to be the predetermined length. It is assumed that the world coordinates at both ends of the edge β1β2 are β1 (x1, y1, 0) and β2 (x2, y2, 0), respectively. The world coordinates of the edge β1β2 are calculated by [Equation 2], [Equation 3], [Equation 4], and [Equation 5].

  Even if the edge is generated on the roof of the large vehicle by the above processing, the position of the vehicle can be calculated appropriately. Furthermore, the distance of the edge β1β2 can be estimated as the vehicle width of the vehicle, and the height of the edge β1β2 can be estimated as the vehicle height.

  FIG. 12 is a process flow diagram of the vehicle tracking means 16. The screen coordinates of the vehicle position are input from the preceding vehicle position estimating means 15 (st121). Based on the screen coordinates, an image of the rear region of the vehicle is cut out from the image and used as an initial template (st122). The size of the template can be determined in proportion to the width of the lane at the cutout position of the image. Pattern collation is performed on the image input next to the frame from which the template is cut out by normalized correlation calculation (st123). If there is no corresponding area by pattern matching, the tracking is terminated (st124). If there is, the screen coordinates are output (st125), and the area is cut out from the image subjected to pattern matching for the screen coordinates. Thus, the template is updated (st126). After updating the template, the process returns to st123 and the tracking process is repeated.

  The traffic index calculation means 17 uses the camera parameters held by the camera parameter holding means 18 to output the screen coordinates relating to the vehicle position output from the vehicle tracking means 16 on the road surface, and the forward projection of FIG. Convert to coordinate system. A passing lane of the vehicle is specified based on the world coordinates of the vehicle, and the speed is measured from the moving distance in a predetermined time.

  By the processing described above, for example, even when a large vehicle is mixed in the image area, the correct vehicle position can be detected without erroneously recognizing the traveling lane where the vehicle exists. Even if it is high, the vehicle can be monitored accurately.

  In this embodiment, as a method of converting the edge from the camera coordinate system to the world coordinate system by the vehicle width conversion means, a method of projecting onto a road surface having a height of zero has been described. However, a plane to be projected is not necessarily a height of zero. There is no need to be on the road surface. For example, the same effect as in the present application can be obtained in a method of projecting onto a plane parallel to the road surface of an arbitrary height.

  Since the present invention can detect the vehicle position with high accuracy by estimating the height and position of a large vehicle without using a plurality of cameras for stereo shooting, it is possible to identify the traffic lane of the large vehicle and the accuracy of speed measurement. Can be improved. It is useful as a traffic information measurement sensor.

The figure which shows an example of implementation of this invention. The figure which shows an example of a vehicle rear part detection means. The figure which shows the example of the process screen of a vehicle rear part narrowing means. The figure which shows the example of a horizontal edge detection. The figure which shows the pinhole model of a street camera. The figure which shows the example of a road overhead image by a road camera. The figure which shows the perspective transformation to a road surface, and the extracted horizontal edge. The figure which shows the relationship between a world coordinate system and a screen coordinate system. The figure which shows conversion to the world coordinate system of the edge by a vehicle width conversion means. The figure which shows the processing flow of a vehicle position estimation means. The conceptual diagram of the pull-back estimation process of a vehicle position. The figure which shows the processing flow of a vehicle tracking means. The figure which shows the example of a Sobel filter.

Explanation of symbols

11 Image input means 12 Vehicle rear narrowing means 13 Horizontal edge detecting means 14 Vehicle width converting means 15 Vehicle position estimating means 16 Vehicle tracking means 17 Traffic index calculating means 18 Camera parameter holding means 21 Reference pattern holding means 22 Pattern collating means 23 Similarity Buffer memory 24 Memory address management means 25 Vehicle rear candidate cutout means 31, 32, 33 Collation area 34, 35, 36 Area 41, 42, 43 Horizontal edge

Claims (16)

In a vehicle monitoring device that monitors a vehicle traveling on the road using an image captured by a camera,
Edge detection means for detecting the edge of the vehicle from the image captured by the camera;
Projecting means for projecting the edge of the vehicle detected by the edge detecting means onto an arbitrary plane;
An edge position that estimates a projection position at which the edge length is equal to or less than the threshold value when the edge length projected by the projection unit is greater than a predetermined threshold value. A vehicle monitoring apparatus comprising: an estimation unit. In claim 1,
A vehicle monitoring apparatus, wherein the height of an edge from the road surface at the edge position estimated by the edge position estimating means is estimated as the height of the vehicle. In claim 1,
A road surface luminance pattern holding unit that holds a luminance pattern of the road surface of the imaging region of the camera;
A vehicle monitoring apparatus, wherein an edge of a vehicle is extracted from an image when a similarity degree between a luminance pattern of a captured image of the camera and a luminance pattern of the road surface is changed from a low state to a high state. In claim 1,
The vehicle monitoring apparatus characterized by estimating the edge position estimated by the said edge position estimation means as a travel lane of a vehicle. In claim 1,
A vehicle monitoring apparatus that measures the speed of a vehicle from a moving distance at a predetermined time of the position of the edge at a predetermined time. In a vehicle monitoring device that photographs a vehicle traveling on a road with a camera and monitors the vehicle using the captured image,
Vehicle edge detection means for detecting the edge of the vehicle from the image taken by the camera;
Projection conversion means for projecting and converting an image captured by the camera onto a road surface according to camera parameters;
First estimation for estimating the position of the edge on the road surface as the vehicle presence position when the length of the vehicle edge of the image projected on the road surface by the projection conversion means is smaller than a predetermined threshold value. Means,
When the length of the vehicle edge of the image projected on the road surface by the projection conversion means is larger than a predetermined threshold value, the position where the edge is projected is pulled back in the direction of the line of sight of the camera, and the edge is A vehicle monitoring apparatus comprising: a second estimation unit configured to estimate a projection position that is equal to or less than the threshold value as a vehicle presence position. In claim 6,
The vehicle monitoring apparatus characterized by estimating the height from the road surface of the vehicle edge in the presence position of the vehicle estimated by the said 2nd estimation means as the height of the said vehicle. In claim 6,
When the length of the vehicle edge of the image projected on the road surface by the projection conversion means is larger than a predetermined threshold value,
The vehicle presence position is calculated as coordinates obtained by multiplying the coefficient obtained by dividing the threshold value by the length of the vehicle edge of the image projected on the road surface by the projection conversion unit. Vehicle monitoring device. In claim 6,
A vehicle monitoring apparatus that estimates the length of a vehicle edge at the vehicle location estimated by the first estimation unit or the second estimation unit as a width of the vehicle. In claim 6,
A collation area for measuring the passing timing of the vehicle in the imaging area of the camera;
A road surface brightness pattern holding unit that holds the brightness pattern of the road surface of the verification region,
A vehicle monitoring apparatus, wherein a vehicle edge is extracted from an image when a similarity degree between a luminance pattern of a captured image of the camera and a luminance pattern of the road surface is changed from a low state to a high state. In claim 6,
A vehicle monitoring apparatus that measures the speed of a vehicle from a moving distance of a vehicle location estimated by the first estimating means or the second estimating means at a predetermined time. In a vehicle monitoring method for monitoring a vehicle traveling on a road by photographing with a camera,
Detecting the edge of the vehicle from the image taken by the camera,
Project the detected edge of the vehicle onto a plane parallel to the road surface,
Vehicle monitoring characterized in that, when the length of the projected edge is larger than a predetermined threshold value, a projection position where the edge length becomes the threshold value is estimated as the existence position of the edge. Method. In claim 12,
A vehicle monitoring method, wherein the height of the edge from the road surface at the estimated position of the edge is estimated as the height of the vehicle. In claim 12,
A vehicle monitoring method, wherein an edge of a vehicle is extracted from an image when a similarity between a luminance pattern of a captured image of the camera and a luminance pattern of the road surface is changed from a low state to a high state. In claim 12,
A vehicle monitoring method, wherein the presence position of the edge is estimated as a travel lane of the vehicle. In claim 12,
A vehicle monitoring method, comprising: measuring a vehicle speed from a transition distance of the edge location at a predetermined time.

Images