JP2001331788A - Vehicle detector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To calculate a distance to a vehicle by precisely specifying the range of the same vehicle even when plural vehicles exist within a prescribed distance range. SOLUTION: This vehicle detector is provided with a distance distribution calculating means 10 which divides image signals image-formed on a pair of image sensors 3 and 4 through a pair of lenses 1 and 2 is divided in the state of meshes, detects the deviation of both of the image signals by each mesh to measure a distance by each mesh based on the principle of a trigonometrical survey, an equal distance cluster extracting means 11 for connecting the meshes at a distance within a prescribed range to extract an equal distance cluster, a vehicle judging means 13 for detecting vehicle judging information constituting the same vehicle such as the horizontal edge of the vehicle, and a vehicle specifying means 14 for connecting the equal distance cluster based on the judging information to specify an area and a distance where the same vehicle exists.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a vehicle detection device which calculates a distance distribution by an optical system using an image sensor and specifies an area where a vehicle exists and a distance from the distance distribution.

[0002]

2. Description of the Related Art Conventionally, three-dimensional measurement techniques using stereo images have been applied in various fields. For example, "3D automatic analysis of stereo images (Information Processing Society of Japan Vol.22,
No.9 pp.846-855, Sep. 1981) ”describes a technique for analyzing and applying stereo photography with a computer. Here, the principle of triangulation is known as a principle of obtaining a distance of a desired point on an image from left and right images having parallax. Then, by obtaining the distance for each of the meshes finely divided on the image by the principle of triangulation, the distance distribution of the object shown in the image can be obtained.

FIG. 12 is a block diagram showing a conventional vehicle detection device using the above-mentioned stereo image. The lenses 1 and 2 are arranged apart from each other by a base length L, and constitute a pair of optical systems, that is, a stereo camera. Image sensors 3 and 4
Is disposed at the position of the focal length f of the lenses 1 and 2, and images captured by the lenses 1 and 2 are individually formed. Analog / digital converters (A / D) 6 and 7 convert the analog output signals of the image sensors 3 and 4 into digital signals. The memories 8 and 9 store the digital output signals of the analog / digital converters 6 and 7, respectively. The distance distribution calculating means 10 is a signal processing device composed of a microcomputer or the like, and has a role of calculating the distance distribution of the preceding vehicle 5 imaged on the lens from the contents of the memories 8 and 9.

Next, the processing operation of the vehicle detection device shown in FIG. 12 will be described. First, the image of the preceding vehicle 5 is individually formed on the image sensors 3 and 4 via the lenses 1 and 2 disposed apart from each other by the base line length L. The image signals of the image sensors 3 and 4 are digitally converted by A / D converters 6 and 7 and stored in memories 8 and 9. The distance distribution calculation means 10 superimposes one image in the memories 8 and 9 on the other image while sequentially shifting the other image in the horizontal direction in pixel units. Then, a distance R from the shift amount S when the images match best to the object (preceding vehicle 5) is obtained by the following equation (A1) based on the principle of triangulation.

R = (f × L) / S (A1)

Where f is the focal length of the lenses 1 and 2,
L is the base line length of the lenses 1 and 2, and the shift amount S is obtained by the number of shift pixels × pixel pitch.

On the other hand, an apparatus for recognizing an obstacle in front of a vehicle based on a distance distribution is disclosed in Japanese Patent Laid-Open No. 9-79821. FIG. 9 is a block diagram showing a configuration of the obstacle recognition device. The obstacle recognition device includes a pair of CCD sensors 1
A calculation area data extracting means 21 for extracting data of a predetermined calculation area based on the respective image data picked up by the CCD sensors 9 and 20;
And comparison area 22 for each calculation area, and a distance calculation means 23 for measuring the distance to the target obstacle for each comparison data extracted by the comparison means 22. Then, the distance calculating means 23
Is a characteristic determining means 24 for determining characteristics such as a distance and a shape of an obstacle in front of the own vehicle based on block information indicating the number and shape of the calculation areas calculated by the means 23.
have. In addition, the feature determination unit 24 includes a calculation area group blocking unit that blocks an unmeasured calculation area close to the calculation area whose distance has been measured, together with the calculation area whose distance has been measured.

That is, in the obstacle recognition device shown in FIG. 9, a calculation area data extracting means 21 and a comparing means 2 are obtained from images obtained by a pair of CCD sensors 19 and 20.
2. A distance distribution for each calculation area as shown in FIG. And the characteristic determination means 2
The vehicle is identified as shown in FIG. 11 by blocking the unmeasured calculation area close to the calculation area whose distance has been measured by 4 together with the calculation area whose distance has been measured. Here, the numerical value of each cell (calculation area) of the mesh in FIG. 10 represents the distance in the mesh (calculation area), and when a distance distribution as shown in FIG. 10 is obtained, the area 27 in FIG. Is determined as shown in FIG.

[0009]

In the above-described conventional vehicle detection device, the determination as to whether or not the objects are the same is made based only on the interval between calculation areas. There was a possibility of misrecognition as.
For example, the left region 25a and the right region 25b in FIG.
Are different objects, there is a problem in that the area 27 is recognized as the same vehicle based on the width of the area 26 between them.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problem, and even when a plurality of vehicles exist within a predetermined distance range, the range of the same vehicle can be accurately specified.
Provided is a vehicle detection device capable of calculating a distance to a vehicle.

[0011]

According to a first aspect of the present invention, an image signal formed on a pair of image sensors via a pair of optical systems is divided into a mesh shape, and each of the mesh signals is divided for each mesh. Distance distribution calculating means for detecting a deviation of an image signal and measuring a distance for each mesh based on the principle of triangulation, and an equidistant cluster for extracting meshes in which the distances are connected to meshes within a predetermined range. Extraction means, vehicle determination means for detecting vehicle determination information constituting the same vehicle, and vehicle identification means for identifying an area and a distance where the same vehicle exists by connecting equidistant clusters based on the determination information. Equipped.

That is, according to the first aspect of the present invention, equidistant clusters obtained from the distance distribution are extracted as candidates for constituting the same vehicle, and when there is a gap between the equidistant clusters, they are extracted as representing characteristics of the same vehicle. By connecting the equidistant clusters using the obtained vehicle determination information, it becomes possible to specify the area and the distance where the same vehicle exists.

According to a second aspect of the present invention, in the first aspect of the present invention, the vehicle determination means extracts an edge as vehicle determination information representing a feature of the same vehicle. Then, when the edge exists between the equidistant clusters, the vehicle specifying unit connects the equidistant clusters.

According to a third aspect of the present invention, in the second aspect of the present invention, an edge of the pair of optical systems in the same direction as the base line length direction is detected. For example, when the base length direction of a pair of optical systems is oriented in the horizontal direction, a horizontal edge (horizontal edge)
To be detected. Then, when this edge exists between the equidistant clusters, the equidistant clusters are connected.

According to a fourth aspect of the present invention, in the first or second aspect of the present invention, the equidistant clusters constituting the same vehicle are determined based on the horizontality or the number of the edges of the optical system extracted by the vehicle determining means with respect to the base line length direction of the optical system. Perform consolidation.

According to a fifth aspect of the present invention, in the first aspect of the present invention, the vehicle determining means extracts a region having a uniform luminance as vehicle determining information, and the vehicle specifying means determines that the region having a uniform luminance exists between the equidistant clusters. Concatenate equidistant clusters.

According to a sixth aspect of the present invention, in the first aspect of the invention, the vehicle judging means extracts a region having a uniform color as the vehicle judging information, and the vehicle specifying means determines that the region having the uniform color exists between the equidistant clusters. Concatenate equidistant clusters.

According to a seventh aspect of the present invention, an area and a distance where a vehicle exists are specified for a plurality of vehicles.

[0019]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiment 1 FIG. 1 is a block diagram showing a configuration of a vehicle detection device according to the present invention. FIG.
, The lenses 1 and 2 are arranged on the vehicle at a distance of the base line length L to form a pair of optical systems, that is, a stereo camera. The image sensors 3 and 4 are arranged at the focal length f of the lenses 1 and 2, respectively.
Is a two-dimensional sensor in which images captured by are formed individually. Analog-to-digital converters (A / D) 6 and 7
Converts the output signals of the image sensors 3 and 4 into digital signals. The memories 8 and 9 include the A / D converter 6
And 7 are stored.

The distance distribution calculating means 10 calculates the distance distribution of the preceding vehicle 5 which is an object imaged on the image sensors 3 and 4 from the contents of the memories 8 and 9. The equidistant cluster extracting means 11 plays a role of extracting an area having a distance within a predetermined range as an equidistant cluster based on the distance distribution calculated by the distance distribution calculating means 10. Memory 12 is A / D
The digital image signal of one of the conversion units 6 and 7 is stored, and the vehicle determination unit 13 calculates information necessary for vehicle determination from the image stored in the memory 12. The vehicle specifying means 14 calculates the area where the vehicle exists and the distance from the information of the distance distribution calculating means 10, the equidistant class extracting means 11 and the vehicle determining means 13.

Next, the operation of the vehicle detecting device according to the first embodiment will be described. In the following description, it is assumed that the lenses 1 and 2, which are a set of optical systems, are mounted on the front part of the vehicle, and are placed so that the base line length direction is in the horizontal direction.

First, the distance distribution of the object (preceding vehicle 5) is
Distance distribution calculation means 1 based on triangulation of stereo images
It is obtained as follows from 0.

That is, the image of the preceding vehicle 5 is individually formed on the image sensors 3 and 4 via the lenses 1 and 2 arranged at a distance of the base line L. The image signals of the image sensors 3 and 4 are digitally converted by A / D converters 6 and 7 and stored in memories 8 and 9.

Here, the image data stored in one memory is divided into mesh areas. 1 of this mesh
Focusing on one, the distance R to the object (preceding vehicle 5) reflected on the mesh is obtained by the following equation (A) using the principle of triangulation.

R = (f × L) / (n × P) (A)

Here, P is a pixel pitch in the base line length direction of the image sensor. At this time, n means that the image of the target mesh is n pixels away from the memories 8 and 9. That is, the parallax in the video imaged by the lenses 1 and 2 is n × P. As a specific method for obtaining n, a correlation method can be used. First, a target mesh is determined based on one memory 8, and a region at the same position as the mesh is selected in the other memory 9. Then, the data in the corresponding areas of the memories 8 and 9 are read, and the sum of the absolute difference values of the data at the corresponding positions of the memories 8 and 9 is calculated. Then, while sequentially shifting the area of the memory 9 one pixel at a time in the base line length direction, the sum of the absolute values of the differences from the data of the target mesh in the memory 8 is obtained in the same manner. The shift amount when the sum of the absolute values of the differences becomes minimum is n. The distance distribution is obtained by similarly calculating the distance for each of the meshes. Note that each mesh area may include an overlapping portion.

Next, the equidistant cluster extracting means 11
Based on the distance distribution calculated by the distance distribution calculating means 10,
An area whose distance is within a predetermined range is extracted as an equidistant cluster. Here, when the base line length direction of the lenses 1 and 2 is the horizontal line direction, an accurate distance tends to be obtained in a region where a vertical edge exists. Therefore, cluster extraction is performed on the preceding vehicle image 5a shown in FIG. In this case, the equidistant clusters extracted by the equidistant cluster extraction unit 11 are divided into the regions 15a, 15b, and 15 shown in FIG.
c, 15d. That is, there is a tendency that a cluster including a vertical edge region and a vertical edge region inside the vehicle, for example, a vertical plate region of a license plate in the outline of the vehicle is easily extracted as equidistant clusters.

Here, the extracted equidistant clusters are:
It does not cover all areas where vehicles exist.
Typically, a gap occurs between clusters as shown in FIG. Therefore, in order to identify the area where the vehicle exists, only the clusters that constitute the same vehicle are connected,
In the case of clusters other than the same vehicle, it is necessary not to connect.

Hereinafter, means for connecting equidistant clusters constituting a vehicle will be described. First, vehicle determination means 1
3 is based on the data stored in the memory 12, FIG.
As shown in the figure, the lateral edges 16a, 16b, 16c,
16d, 16e and 16f are extracted by image processing. As the horizontal edge extraction filter in this case, for example, Sob
An el filter can be used. Next, the vehicle specifying means 14 generates the equidistant clusters 15a, 15b, 15c,
15d information and horizontal edges 16a, 16b, 16c, 16
Using the information of d, 16e, and 16f, the area where the vehicle exists is specified. Equidistant clusters 15a, 15b, 1
5c, 15d and horizontal edges 16a, 16b, 16c, 16
FIG. 5 shows a superposition of d, 16e and 16f. As shown in FIG. 5, a horizontal edge exists between equidistant clusters constituting a vehicle. Here, in order to more surely identify that the vehicle is a vehicle, it can be used as a condition that the extracted horizontal edge is horizontal and that there are a plurality of horizontal edges.

If the plurality of equidistant clusters of interest have a high probability of being the same vehicle from the above information, as shown in FIG. 6, the area 17 between equidistant clusters and the equidistant clusters 15a, 15b, 15c , 15d are determined to be present in the area where the vehicle is connected. Then, the distance to the vehicle can be calculated from the distance distribution of the area where the vehicle exists. Here, by calculating the average of the individual distances of the mesh in the area where the vehicle exists from the distance distribution, it is possible to precisely calculate the distance of the preceding vehicle 5 which is the target object.

As described above, according to the first embodiment, the distance distribution for each mesh is obtained by the distance distribution calculating means 10, and meshes whose distances are within a predetermined range are connected by the equidistant cluster extracting means 11. Distance clusters are extracted, and if there is a gap between the equidistant clusters, the equidistant clusters are connected using the edges extracted as representing the characteristics of the same vehicle. Can be accurately specified and the distance can be calculated accurately.

Embodiment 2 FIG. In the first embodiment, a horizontal edge is extracted by image processing, and an equidistant cluster is connected by the horizontal edge to specify an area where the same vehicle exists. In the second embodiment, the luminance is determined in the horizontal direction. Alternatively, an area in which the same vehicle exists is specified by extracting an area having a uniform color and connecting equidistant clusters with the area having the uniform luminance or color. In addition,
The configuration example of the device of the second embodiment is the same as that of FIG. 1 of the first embodiment.

The means for connecting the equidistant clusters constituting the vehicle according to the second embodiment will be described. As shown in FIG. 7, in a vehicle, a region 18 in which the brightness or the color is uniform in the lateral direction.
a, 18b, 18c, 18d, and 18e exist in a plurality of strips. Further, there is a tendency that the distance cannot be calculated by the distance distribution calculating means 10 in these areas. Vehicle determination means 1
In FIG. 7, by searching for a region in which the luminance or the color is uniform in the horizontal direction with respect to the region in which the distance distribution is not calculated, FIG.
Are extracted as uniform regions of brightness or color. In the vehicle specifying means 14, the clusters 15a, 15b,
As shown in FIG. 8, the areas 18a, 18b, 18c, 18d, and 18 having uniform brightness or color are provided between the areas 15c and 15d.
When connected by e, the connected equidistant clusters are specified as the vehicle existence region, as in FIG.

As described above, according to the second embodiment, the distance distribution for each mesh is obtained by the distance distribution calculating means 10, and meshes whose distances are within a predetermined range are connected by the equidistant cluster extracting means 11. Since distance clusters were extracted, and when there was a gap between the equidistant clusters, the equidistant clusters were connected using uniform regions of brightness or color extracted as representing the characteristics of the same vehicle, It is possible to accurately specify the area where the same vehicle exists and calculate the distance.

Other Embodiments In the above-described embodiment, the case where the number of preceding vehicles is one has been described. However, even when there are a plurality of vehicles, the same processing is performed in each of the regions, so that the existence regions of the plurality of vehicles are specified, It is possible to calculate the distance to. Further, in the above-described embodiment, the case where the preceding vehicle is detected has been described. However, the following vehicle traveling behind the host vehicle may be detected.

Further, in the above-described embodiment, the pair of lenses constituting the stereo camera are arranged so that their base lines extend in the horizontal direction of the traveling road, but they are arranged in the direction perpendicular to the traveling road surface. May be.

[0037]

As described above, according to the present invention, the equidistant clusters are extracted from the distance distribution information, and the equidistant clusters are connected by using the vehicle determination information which is the characteristic of the same vehicle, so that the area of the same vehicle is obtained. Therefore, even when a plurality of objects exist within a predetermined distance range, the range of the same vehicle can be accurately specified, and the distance to the vehicle can be calculated reliably.

As the vehicle determination information which is a feature of the same vehicle, the presence of an edge, particularly the presence of an edge perpendicular to the base line direction of a pair of optical systems is detected, and equidistant clusters are connected. It is possible to accurately specify the range of the same vehicle and calculate the distance. Furthermore, by using the information on the number and horizontality of the edge,
More accurate vehicle identification and distance calculation can be performed.

Further, as vehicle determination information which is a feature of the same vehicle, the presence of a region having uniform brightness or color is detected,
Since the equidistant clusters are connected, it is possible to accurately specify the range of the same vehicle and calculate the distance.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a vehicle detection device according to the present invention.

FIG. 2 is a schematic diagram of an image of a preceding vehicle that is a target in the first embodiment.

FIG. 3 is a diagram showing an example of regions of equidistant clusters extracted in the first embodiment.

FIG. 4 is a diagram showing an example of a horizontal edge extracted in the first embodiment.

FIG. 5 is a diagram in which equidistant clusters and horizontal edges are superimposed in the first embodiment.

FIG. 6 is a diagram for describing an area of a vehicle specified in the first embodiment.

FIG. 7 is a diagram showing an example of a region of uniform luminance or color extracted in Embodiment 2 of the present invention.

FIG. 8 is a diagram in which an equidistant cluster and a region with uniform luminance or color are superimposed in Embodiment 2 of the present invention.

FIG. 9 is a block diagram illustrating a configuration of a conventional obstacle recognition device.

FIG. 10 is a diagram illustrating an example of a distance distribution obtained by the obstacle recognition device of FIG. 9;

11 is a diagram illustrating an area of a vehicle specified by the obstacle recognition device of FIG. 9;

FIG. 12 is a configuration diagram showing a conventional vehicle detection device using a stereo image.

[Explanation of symbols]

1,2 lens, 3,4 image sensor, 5 preceding car, 6,7 analog / digital converter, 8,9,1
2 memory, 10 distance distribution calculating means, 11 equidistant cluster extracting means, 13 vehicle determining means, 14 vehicle specifying means, 15a to 15d equidistant cluster, 16a to
16f horizontal edge, 17 connected area, 18a-1
8e Uniform area of brightness or color.

Claims (7)

[Claims] An image signal formed on a pair of image sensors via a pair of optical systems is divided into meshes, and a shift between the two image signals is detected for each mesh, based on the principle of triangulation. Distance distribution calculating means for measuring the distance of each mesh, and equidistant cluster extracting means for extracting equidistant clusters by connecting meshes whose distances are within a predetermined range,
Vehicle determining means for detecting vehicle determining information constituting the same vehicle; and vehicle specifying means for specifying an area and a distance where the same vehicle exists by connecting equidistant clusters based on the determining information. A vehicle detection device characterized by the following. 2. The vehicle determining means extracts edges as vehicle determining information, and the vehicle specifying means connects equidistant clusters when the extracted edges exist between equidistant clusters. Item 4. The vehicle detection device according to item 1. 3. The edge extracted by the vehicle determination means is:
3. The vehicle detection device according to claim 2, wherein the edges are substantially the same direction as the base line length direction of the pair of optical systems. 4. Based on the number of edges extracted by the vehicle determination means or the horizontality of the optical system with respect to the base line length direction,
4. The vehicle identification unit according to claim 2, wherein the vehicle identification unit connects equidistant clusters constituting the same vehicle.
The vehicle detection device according to claim 1. 5. The vehicle judging means extracts a region with uniform luminance as vehicle judgment information, and the vehicle specifying means connects equidistant clusters when the region with uniform luminance exists between equidistant clusters. The vehicle detection device according to claim 1, wherein: 6. The vehicle determining means extracts a uniform color area as vehicle determining information, and the vehicle specifying means connects the equidistant clusters when the uniform color area exists between the equidistant clusters. The vehicle detection device according to claim 1, wherein: 7. The vehicle detection device according to claim 1, wherein an area and a distance where the vehicle exists are specified for a plurality of vehicles.