JPH07129899A - Vehicle recognition device - Google Patents

Abstract

PURPOSE:To provide a vehicle recognition device which properly extracts only a precedent vehicle on the same lane with this vehicle irrespectivily of the travel position of this vehicle and the position relation between this vehicle and other peripheral vehicle in a device which extracts a precedent vehicle from an image obtained by photographing the road front. CONSTITUTION:A vehicle lower end extracting means 5 and a vehicle side end extracting means 6 performs edge search processing in the lane area of this vehicle extracted by using an extracting means 3 for this vehicle and adjacent lane areas extracted by an adjacent lane extracting means 4 as to a road edge image obtained by an image input means 1 and an image differentiating means 2 and a vehicle area judging means 7 cuts both areas of the precedent vehicle. Consequently, only the precedent vehicle on the same lane as the lane of this vehicle can properly by extracted regardless of the travel position of this vehicle, and the positional relation between this vehicle and peripheral vehicles.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to preventive safety technology for automobiles,
In particular, the present invention relates to an apparatus for recognizing a vehicle in front in the lane area in which the vehicle is traveling by using image processing technology.

[0002]

2. Description of the Related Art Conventionally, as a device for recognizing a forward vehicle of this type, as described in, for example, Japanese Patent Application Laid-Open No. 1-281600, an edge component existing in a road region cut out in advance is tracked. There was one that extracts the area where the vehicle in front exists.

[0003]

However, in such a conventional vehicle recognition device, when a front vehicle is changing lanes, or when a plurality of vehicles run in adjacent lanes in parallel with the front vehicle. In such a case, there is the first problem that it is difficult to properly extract only the vehicle ahead.

Further, the conventional vehicle recognition device has a second problem that it is difficult to extract the upper end portion of a forward vehicle having an overlap with another vehicle or various road structures. .

The present invention solves the problems as described above. First, by searching for a vehicle ahead in the lane in which the host vehicle is traveling, the search range is expanded to the adjacent lane. (EN) Provided is a device capable of appropriately extracting only a front vehicle existing in a lane in which an own vehicle is traveling, irrespective of a traveling position of the own vehicle and a relative position between the own vehicle and a surrounding vehicle. It has one purpose.

Further, the present invention provides a device capable of appropriately extracting the upper end portion of a front vehicle even when the front vehicle overlaps with another vehicle, a road wall, an overpass, or the like. It has a secondary purpose.

[0007]

In order to achieve the above first object, the first means of the present invention is to provide an image input means for forward photographing mounted on a vehicle, and a road input by the image input means. An image differentiating means for generating an edge image by applying a differentiating process to the image, a own lane extracting means for extracting a lane area in which the own vehicle is traveling from the edge image generated by the image differentiating means, and the image differentiating means. Adjacent lane extracting means for extracting a lane area adjacent to the traveling lane of the own vehicle extracted by the own lane extracting means from the edge image generated by the own lane extracting means, and the own vehicle extracted by the own lane extracting means. Vehicle lower end extraction means for extracting the lower end of the forward vehicle candidate area from the edge image generated by the image differentiation means in the lane area in which the vehicle is traveling, and the adjacent vehicle The left and right edges of the forward vehicle candidate area are extracted from the edge image generated by the image differentiating means within the area defined by the adjacent lane extracted by the extracting means and the lower end of the vehicle candidate area extracted by the vehicle lower edge extracting means. The vehicle side edge extraction means, the area cut out by the lower edge of the forward vehicle candidate area extracted by the vehicle lower edge extraction means and the left and right edges of the forward vehicle candidate area extracted by the vehicle side edge extraction means correspond to the forward vehicle. Vehicle area determining means for determining whether or not to do so.

Further, in order to achieve the above first and second objects, the second means of the present invention, in addition to the structure of the above first means, is judged by the vehicle area judging means to be a forward vehicle. In the vehicle area, the vehicle model adaptation means is provided for cutting out the vehicle area by applying the vehicle model to the edge image generated by the image differentiating means.

In order to achieve the above first and second objects, the third means of the present invention, in addition to the structure of the above first means, is judged by the vehicle area judging means to be a forward vehicle. In the vehicle area, there is provided rear window extracting means for cutting out the vehicle area by extracting the rear window of the front vehicle from the edge image generated by the image differentiating means.

[0010]

With the first means described above, after the preceding vehicle is searched in the lane in which the host vehicle is traveling, the search range can be expanded to the adjacent lane. Regardless of the position and the relative positions of the own vehicle and the surrounding vehicles, only the forward vehicle existing in the lane in which the own vehicle is traveling can be appropriately extracted.

Further, by the second and third means, the upper end of the vehicle is extracted by using the vehicle model, the rear window model, etc., so as to prevent the vehicle from overlapping with other vehicles and various road structures. Even in the case of holding, only the vehicle in front can be appropriately extracted.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the basic configuration of a vehicle recognition device according to the first embodiment of the present invention.

Reference numeral 1 denotes an image input means for photographing the front of the vehicle, which is realized by installing a CCD camera at the front of the vehicle so as not to obstruct the driver's view, as shown in FIG. Reference numeral 2 denotes an image differentiating means, which differentiates the road image input from the image inputting means 1 to generate an edge image. Reference numeral 3 denotes own lane extracting means for extracting the lane area in which the own vehicle is traveling from the edge image generated by the image differentiating means 2, which is realized by linearly approximating the outline point sequence of the white line. 4 is own lane extracting means 3
It is an adjacent lane extracting means for extracting a lane area adjacent to the own lane extracted by the above, and performs a process of enlarging the own lane area. Reference numeral 5 denotes a vehicle lower end extraction means, which extracts the lower end of the forward vehicle candidate area by performing an edge search within the lane area in which the own vehicle is running extracted by the own lane extraction means 3. Reference numeral 6 denotes a vehicle side edge extraction means, which extracts the left and right edges of the forward vehicle candidate area by performing an edge search within the adjacent lane area extracted by the adjacent lane extraction means 4. Reference numeral 7 denotes a vehicle area determination means, which is a vehicle lower end extraction means 5 and a vehicle side edge extraction means 6
It is to determine whether or not the front vehicle candidate area cut out by corresponds to the front vehicle.

Next, the operation of the vehicle recognizing device according to the first embodiment of the present invention, which comprises the above-mentioned components, will be described with reference to the flow chart shown in FIG. First step 1
In 01, the front of the vehicle is photographed using the image input means 1,
Import road image data. FIG. 7 shows an example of the captured image.

In step 102, the image differentiating means 2 is used to extract edge components from the road image captured in step 101. A differential operator such as Sobel is used for edge extraction, and the obtained edge strength is binarized with an appropriate threshold value to generate an edge image. FIG. 8 shows an edge image created from the input image of FIG.

In step 103, the own lane extracting means 3 is used to extract the lane region in which the own vehicle is traveling. The extraction of the lane area in which the host vehicle is traveling is performed in step 102.
The outline candidate point sequence of the white line is extracted from the edge image generated in (1), and these point sequences are linearly approximated. In this case, the processing range is limited to the lower half of the image to improve the accuracy of the linear approximation processing. FIG. 4 shows a flow of this processing. First, in step 201, processing for extracting contour points of the left and right white lines is performed. First, as shown in FIG. 9, the scanning line is searched leftward from the center I of the edge image, and the first pixel whose edge strength E (x, y) exceeds the threshold value The is the contour point inside the left white line. To do. Similarly, the scanning line is searched for from the center I of the image shown in FIG. 9 to the right, and the first pixel whose edge strength E (x, y) exceeds the threshold value The is defined as the contour point inside the left white line. To do. FIG. 10 shows the result of extracting the contour point sequences of the left and right white lines. Next, in step 202,
A process of directly approximating the contour point sequences of the left and right white lines extracted in step 201 is performed. Note that Houg is used for the linear approximation processing.
The method of h conversion is used. Thus, as shown in FIG.
Triangle VLR formed by the straight line approximation result of the left white line on the road surface, the straight line approximation result of the right white line, and the lower edge of the image
Is extracted as the own lane area in which the own vehicle is traveling.

Next, in step 104, the road area including the adjacent lane is approximately extracted by the adjacent lane extracting means 4 based on the lane area in which the own vehicle is traveling thus extracted. In this case, as shown in FIG. 12, a triangle VL'R 'obtained by extending the length of the bottom side LR of the triangle VLR representing the own lane area to the left and right is extracted as a road area including adjacent lanes.

Next, in step 105, the vehicle lower end extraction means 5 is used to set the lower end of the candidate region of the preceding vehicle and the center of gravity in the scanning line direction within the own lane region in which the own vehicle is running obtained in step 103. Extract the position. In Figure 5,
A series of flow of this processing is shown. First, in step 301, the number of edge points existing on each scanning line in the lane area in which the vehicle is traveling is counted, and a histogram as shown in FIG. 13 is created. In addition, step 302,
As indicated by 303, the histogram is searched from the lower side, and the scanning line position where the number of edge points exceeds the threshold value EB for the first time is extracted as the lower end of the forward vehicle candidate area. Then, the process proceeds to step 304, where the average value of the edge point sequence distributed on the scanning line at the lower end of the forward vehicle candidate region extracted in step 303 is calculated and set as the barycentric position of the forward vehicle candidate region in the scanning line direction. Conversely, in step 302, the threshold value E
If there is no scanning line exceeding B, it is determined that there is no vehicle in front as shown in step 305, and the process ends.

In step 106 of FIG. 3, the vehicle side edge extraction means 6 is used to expand the processing area to a road area including an adjacent lane, and the left and right edges of the forward vehicle candidate area are extracted. At this time, the lower end position of the forward vehicle candidate region extracted by the vehicle lower end extraction unit 5 is set as the lower limit of the processing region. A series of flow of this process is shown in FIG. First, in step 401, the number of edge points existing on each pixel column in the direction perpendicular to the scanning line is counted, and a histogram as shown in FIG. 14 is created. Next, in steps 402 and 403, the frequency of the histogram is examined from the center of gravity in the scanning line direction of the forward vehicle candidate area extracted by the vehicle lower edge extraction means 5 to the left, and the first position where the edge score becomes equal to or less than the threshold value ES is determined. It is extracted as the left end of the forward vehicle candidate area. Similarly, in steps 404 and 405, the histogram frequency is examined from the center of gravity position to the right, and the first position where the edge score becomes equal to or less than the threshold value ES is extracted as the right side of the forward vehicle candidate region. On the contrary, if the position that is equal to or less than the threshold value ES is not detected in step 402 or step 404, it is determined in step 406 that there is no vehicle in front, and the process ends.

In step 107, the vehicle lower end extraction means 5
With respect to the front vehicle candidate area cut out by the vehicle side edge extraction means 6, it is determined whether or not the area corresponds to a front vehicle. In this case, the number of horizontal edges is equal to the threshold value EH in the recessed area shown in FIG.
When it exceeds, the front vehicle candidate area cut out by the vehicle lower end extraction means 5 and the vehicle side end extraction means 6 is
It is determined that the vehicle is a forward vehicle existing in the lane area in which the vehicle is traveling. On the contrary, the number of horizontal edges is the threshold value EH.
In the following cases, it is determined that the cut out candidate area is a shadow on the road surface.

In this way, after extracting the lower end of the front vehicle area and the center of gravity in the scanning line direction within the lane area in which the host vehicle is traveling, the front vehicle candidate area within the road area including the adjacent lane is extracted. By searching for edges from the center of gravity in the scanning line direction to the outside on the left and right, even when the vehicle in front is changing lanes or when another vehicle is traveling in the adjacent lane, It is possible to appropriately extract only the forward vehicle existing in the traveling lane of the vehicle.

Next, a second embodiment of the present invention will be described with reference to the drawings. FIG. 16 is a block diagram showing the basic configuration of the second embodiment of the present invention. The basic configuration of the second embodiment is the basic configuration of the first embodiment with vehicle model adaptation means 8 added. The vehicle model adaptation means 8 extracts the upper end portion of the vehicle by adapting the vehicle model to the front vehicle area determined by the vehicle area determination means 7, and finally cuts out a rectangular area circumscribing the front vehicle.

Next, the operation of the vehicle recognition apparatus according to the second embodiment of the present invention to which the vehicle model adaptation means 8 is added will be described with reference to the flow chart shown in FIG. It should be noted that the detailed contents are shown only for the parts different from the first embodiment, and the other parts are according to the first embodiment.

First, the processing from steps 501 to 507 corresponds to the processing from step 101 to step 107 in FIG. 3 shown in the first embodiment. The same processing is performed respectively, and the process proceeds to step 508. .

In step 508, the vehicle area determination means 7
By adapting the vehicle model to the area determined to correspond to the front vehicle in (1), the upper end of the front vehicle is extracted and the rectangular area circumscribing the front vehicle is finally cut out. Generally, since the upper end portion of the vehicle ahead overlaps various road structures, it is more difficult to extract than the lower end portion, the left end portion, and the right end portion. Therefore, we prepare rectangular models of various sizes according to the vehicle type such as ordinary passenger cars, light cars, vans, trucks, etc. in advance, and apply this to the edge image to extract the upper end of the front vehicle. Cut out the vehicle area.

Generally, as shown in FIG. 19, the ratio of the total width to the total height of an automobile differs depending on the vehicle type such as a normal passenger car, a light automobile, a van, and a truck. Therefore, the ratio Sn of the total width Wn and the total height Hn of an average ordinary passenger car is
Is defined as the formula (1), and the average width W of the average light vehicle is
The ratio Sk of the k and the total height Hk is defined by the formula (2), the ratio Sv of the average full width Wv of the van and the total height Hv is defined by the formula (3), and the average width Wt of the truck is The ratio St of the total height Ht is defined as in Expression (4). However, Sn, S
It is assumed that k, Sv, and St comply with the equation (5).

Hn / Wn = Sn (1) Hk / Wk = Sk (2) Hv / Wv = Sv (3) Ht / Wt = St (4) Sn <Sk <Sv <St (5) The width on the image of the vehicle ahead is W
Suppose In this case, assuming that the vehicle in front is a normal passenger car, the height H on the image can be expressed by Expression (6). Similarly, assuming that the preceding vehicle is a light vehicle, a van, or a truck, its height H can be expressed as in equations (7), (8), and (9), respectively.

H = Sn × W (6) H = Sk × W (7) H = Sv × W (8) H = St × W (9) First, in step 601, the vehicle side end extraction means 6 extracts the data. The lateral width W of the front vehicle on the image is calculated based on the left and right ends of the front vehicle. Next step 6
In 02, the height H on the image when the vehicle type of the forward vehicle is each of a normal passenger car, a light vehicle, a van, and a truck is calculated according to the equations (6) to (9),
Four types of rectangular models having a width W and a height H are created. Further, in step 603, each of these four types of models is applied to the edge image. In this case, the bottom edge, left edge, and right edge of the rectangular model are fitted so as to match the bottom edge, left edge, and right edge of the front vehicle area cut out by the vehicle area determination means, respectively, and the edge points on the scanning line corresponding to the upper side of the model are Count the number. Further, in step 604, of the four types of rectangular models, the model with the largest number of edge points is determined to be the model that matches the vehicle type of the forward vehicle, and the rectangular area surrounded by this model is determined as the forward vehicle. Finally cut out as the area circumscribing. FIG. 20 shows a result of cutting out a rectangular area circumscribing the preceding vehicle.

As described above, the vehicle model created based on the lower end and the left and right ends of the front vehicle extracted from the road region is applied to extract the vehicle upper end, thereby appropriately cutting out the rectangular region circumscribing the front vehicle. be able to.

Next, a third embodiment of the present invention will be described with reference to the drawings. FIG. 21 is a block diagram showing the basic configuration of the third embodiment of the present invention. The basic configuration of the third embodiment is such that the vehicle model adaptation means 8 in the basic configuration of the second embodiment is replaced with rear window extraction means 9. Generally, the rear window of a passenger vehicle is a vehicle feature that appears relatively prominently in the image, and its upper end portion substantially coincides with the upper end of the vehicle. Furthermore, the shape and size of a passenger car or a light car are almost constant regardless of the vehicle type. The rear window extraction means 9 extracts the rear window of the front vehicle from the rectangular area formed by the lower edge, the left edge, the right edge, and the upper edge of the image of the front vehicle cut out by the vehicle area determination means 7. Extract the upper end of
Finally, the rectangular area circumscribing the vehicle in front is cut out.

Next, the operation of the vehicle recognizing device according to the third embodiment of the present invention having the rear window extracting means 9 will be described with reference to the flow chart shown in FIG. It should be noted that the detailed contents are shown only for the parts different from the second embodiment, and the other parts are according to the second embodiment.

First, the processing from step 701 to 707 corresponds to the processing from step 501 to 507 in FIG. 17 shown in the second embodiment, and the same processing is performed respectively, and the process proceeds to step 708.

In step 708, the upper end portion of the front vehicle is extracted by searching the rear window of the front vehicle using the rear window extraction means 9 and the front vehicle region is cut out. FIG. 23 shows a flow of this processing. First, in step 801, the vehicle side edge extraction means 7
The lateral width W of the front vehicle on the image is calculated from the left and right ends of the front vehicle region extracted in step S3. Then, in step 802, as shown in FIG. 24, a trapezoidal rear window model having an upper bottom W × β, a lower bottom W × γ, and a height W × α is created. In step 803, using this motel as a template, the vehicle is moved within a rectangular area composed of the lower end, the left end, the right end, and the upper end of the image of the forward vehicle shown in FIG. 24, and matching with the edge image is performed. In this case, as a measure showing the degree of conformity between the template and the edge image, a number obtained by counting the number of edge points existing on the trapezoidal contour of the template is used. Further, in step 804, the maximum value of this fitness is calculated, and in step 805 it is judged whether or not this maximum value exceeds the threshold value Mw. If the maximum value of the goodness of fit exceeds the threshold Mw, step 806.
Go to and extract the position of the template with the maximum matching value as the rear window of the vehicle ahead. And
The upper end portion of this rear window is approximated as the upper end position of the vehicle ahead. Further, in step 807, the upper end of the preceding vehicle thus extracted, the vehicle area determination means 7
The rectangular region surrounded by the lower end, the left end, and the right end obtained by is cut out as a region circumscribing the front vehicle. FIG. 20 shows the result of cutting out the front vehicle as a rectangular area. On the contrary, if the maximum value of the conformity is equal to or less than the threshold value Mw, it is determined that the vehicle in front is a vehicle such as a van or a large truck, and in step 808, the vehicle upper end portion is detected by applying the model for the van or truck. Extract. First, the ratio S of the average van width Wv and height Hv
v, the ratio St of the total width Wt and the total height Ht of the average track is defined as in equations (3), (4), and (5), respectively.

Hv / Wv = Sv (3) Ht / Wt = St (4) Sv <St (5) Further, using the lateral width W of the front vehicle on the image calculated in step 801, this front vehicle is a van. Assuming that there is, the height H on the image can be expressed as in equation (8), and if it is assumed to be a track, in equation (9).

H = Sv × W (8) H = St × W (9) Therefore, two types of rectangular models having the width W and the height H thus calculated are created, and each of them is an edge image. Apply to. In this case, the lower end, the left end, and the right end of the rectangular model are fitted so as to match the lower end, the left end, and the right end of the front vehicle region cut out by the vehicle region determination means 7, respectively, and the edge points on the scanning line corresponding to the upper side of the model are fitted. To count the number of.

Further, among these two types of rectangular models,
The model with the larger number of edge points is determined to be a model suitable for vehicle identification of the front vehicle, and the rectangular area surrounded by this model is finally cut out as an area circumscribing the front vehicle.

In this way, by using the rear window as the feature quantity, even when the vehicle in front has an overlap with other vehicles and various road structures,
The front vehicle area can be appropriately cut out.

The vehicle lane area extracting means 3 of the above embodiment
Although the Hough transform is used for the linear approximation processing of the white line in, the linear approximation method such as the least square method may be used. Further, the Sobel operator is used to extract the edge component in the image differentiating means 2 of the above embodiment, but other edge detection operators such as Laplacian may be used. Further, the vehicle model adaptation means 8 may be one that extracts the contour of the vehicle using a method such as an active contour model.

Each means of the present invention can be realized by software using a computer, or can be realized by using a dedicated hardware circuit having each of these functions.

[0040]

As described above, according to the present invention, the own lane extracting means for extracting the lane area in which the own vehicle is traveling, the adjacent lane extracting means for extracting the adjacent lane area, and the own vehicle traveling lane. Vehicle lower end extraction means for extracting the lower end of the forward vehicle candidate region from the edge image in the region, and vehicle for extracting the left and right ends of the forward vehicle candidate region from the edge image within the region defined by the adjacent lane and the lower end of the vehicle candidate region First, by providing the side edge extraction means and the vehicle area determination means for determining whether the area cut out by the lower end of the front vehicle candidate area and the left and right edges of the front vehicle candidate area corresponds to the front vehicle, After searching for a vehicle in front of you in the lane you are driving, expand the search range to the adjacent lane to determine the location of your vehicle and the relative position of your vehicle and surrounding vehicles. Without only forward vehicle present in the lane in which the vehicle traveling can be properly extracted.

Vehicle model adaptation means for cutting out the vehicle region by adapting the vehicle model to the edge image,
A rear window extraction unit that cuts out the vehicle region by extracting the rear window of the vehicle ahead from the edge image is provided, and the upper end portion of the vehicle is extracted by using the vehicle model or the rear window model. Even when there is an overlap with a vehicle or various road structures, only the vehicle in front can be appropriately extracted.

[Brief description of drawings]

FIG. 1 is a block diagram showing a basic configuration of a vehicle recognition device according to a first embodiment of the present invention.

FIG. 2 is a plan view showing an example of mounting the camera on the vehicle of the vehicle recognition device.

FIG. 3 is a flowchart showing an operation of the vehicle recognition device.

FIG. 4 is a flowchart showing an operation of own lane area extraction processing of the vehicle recognition device.

FIG. 5 is a flowchart showing an operation of vehicle region lower end extraction processing of the vehicle recognition device.

FIG. 6 is a flowchart showing an operation of vehicle side edge extraction processing of the vehicle recognition device.

FIG. 7 is a diagram showing an example of an input image in the vehicle recognition device.

FIG. 8 is a diagram showing an edge image created from an input image in the vehicle recognition device.

FIG. 9 is a diagram showing a white line contour extraction process in the vehicle recognition device.

FIG. 10 is a diagram showing a white line contour extraction result in the vehicle recognition device.

FIG. 11 is a diagram showing an extraction result of a lane region in which the vehicle is traveling in the vehicle recognition device.

FIG. 12 is a diagram showing an extraction process of an adjacent lane area in the vehicle recognition device.

FIG. 13 is a diagram showing a lower end extraction process of a front vehicle in the vehicle recognition device.

FIG. 14 is a diagram showing a process of extracting left and right ends of a front vehicle in the vehicle recognition device.

FIG. 15 is a diagram showing a result of extracting a forward vehicle candidate region in the vehicle recognition device.

FIG. 16 is a block diagram showing a basic configuration of a vehicle recognition device according to a second embodiment of the present invention.

FIG. 17 is a flowchart showing the operation of the vehicle recognition device.

FIG. 18 is a flowchart showing an operation of vehicle model adaptation processing of the vehicle recognition device.

FIG. 19 is a diagram showing rectangular models for various vehicle types.

FIG. 20 is a view showing a result of extracting a rectangular area circumscribing a vehicle ahead of the vehicle recognition device.

FIG. 21 is a block diagram showing a basic configuration of a vehicle recognition device according to a third embodiment of the present invention.

FIG. 22 is a flowchart showing the operation of the vehicle recognition device.

FIG. 23 is a flowchart showing the operation of the rear window extraction process of the vehicle recognition device.

[Figure 24] Diagram showing the model of the rear window

[Explanation of symbols]

 1 Image Input Means 2 Image Differentiating Means 3 Own Lane Extracting Means 4 Adjacent Lane Extracting Means 5 Vehicle Lower Edge Extracting Means 6 Vehicle Side Edge Extracting Means 7 Vehicle Area Determining Means 8 Vehicle Model Adapting Means 9 Rear Window Extracting Means

Claims (3)

[Claims] 1. An image input means for forward photography mounted on a vehicle, an image differentiating means for generating an edge image by subjecting a road image input by said image input means to differential processing, and said image differentiating means. Own lane extracting means for extracting the lane area in which the own vehicle is traveling from the generated edge image, and running of the own vehicle extracted by the own lane extracting means from the edge image generated by the image differentiating means. An adjacent lane extracting means for extracting a lane area adjacent to an existing lane, and a forward vehicle candidate from the edge image generated by the image differentiating means in the lane area in which the own vehicle extracted by the own lane extracting means is traveling. The vehicle lower edge extracting means for extracting the lower edge of the region, the adjacent lane extracted by the adjacent lane extracting means, and the vehicle lower edge extracting means. The vehicle side edge extracting means for extracting the left and right edges of the forward vehicle candidate area from the edge image generated by the image differentiating means in the area defined by the lower edge of the vehicle candidate area extracted by the vehicle lower edge extracting means. Vehicle region determining means for determining whether or not a region cut out by the lower end of the extracted forward vehicle candidate region and the left and right ends of the forward vehicle candidate region extracted by the vehicle side edge extracting means corresponds to a forward vehicle. Vehicle recognition device. 2. A vehicle model adaptation means for cutting out a vehicle area by applying a vehicle model to an edge image generated by an image differentiating means in a vehicle area determined to correspond to a forward vehicle by the vehicle area determination means. Item 1. The vehicle recognition device according to item 1. 3. A rear area cut out from the vehicle area by extracting a rear window of the front vehicle from the edge image generated by the image differentiating means in the vehicle area determined to correspond to the front vehicle by the vehicle area determining means.
The vehicle recognition device according to claim 1, further comprising window extraction means.