JP3436224B2 - Vehicle end detection device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a vehicle end detection device for detecting a vehicle end which is an upper end or a lower end of a vehicle region on an image as information such as a width and a position of a vehicle in a traffic flow measuring system.

[0002]

2. Description of the Related Art Intelligent Transport System (IT
In the field of S)), a traffic flow measurement system that measures the number of vehicles, the vehicle speed, or the like from an image captured by a camera is important.

As an apparatus for detecting the end of a vehicle in order to measure the number of vehicles or the speed of the vehicle, for example, Japanese Patent Laid-Open No. 282/1982
Japanese Patent No. 593 discloses a vehicle head detecting device for detecting the head of a running vehicle (prior art example 1). The vehicle head detection device described in the conventional example 1 photographs a predetermined area on the road with a fixed camera, obtains a contour image from the image, scans the obtained contour image in the horizontal direction, counts the number of contour pixels, and vertically. Ask for a profile. Then, after a predetermined time has elapsed from the shooting time of the image corresponding to this vertical profile, the vertical profile is obtained in the same manner, and from the difference profile that is the difference between the pair of vertical profiles thus obtained, the white line drawn on the road is fixed. It seeks the position of the vehicle head without being affected by foreign matter.

Further, Japanese Unexamined Patent Publication No. 7-28975 discloses a road environment recognition device which obtains the lowermost portion of a vehicle ahead of a traveling road from the image of a vehicle-mounted camera (conventional example 2).
The road environment recognition device according to the second conventional example performs smoothing on an input image to make it difficult to detect an image having a large relative speed with a camera, and further extracts horizontal edges to generate a horizontal edge image.

Next, the generated horizontal edge image is sequentially scanned from the bottom to scan the horizontal line to count the horizontal edge pixels on the line, and the horizontal edge pixels equal to or more than a preset number of pixels are included. If there is such a horizontal line, it will be the candidate line at the bottom of the vehicle ahead. Furthermore,
A scanning area with a predetermined height is set above the screen from the detected candidate lines, the candidate lines included in that area are detected, and the number of them is counted. The candidate line at the bottom of the exceeded scanning area is set as the bottom of the preceding vehicle.

[0006]

However, the vehicle head detecting device described in the conventional example 1 can detect only the vehicle end of the frontmost one vehicle in the screen, and therefore, when a plurality of vehicles are photographed. However, there is a problem that only the vehicle end of one vehicle can be detected.

In addition, the road environment recognition device described in the second conventional example aims to detect the lower end of the vehicle closest to the camera, and to individually determine the lower end of the vehicle when there are a plurality of vehicles. Since it is not assumed, when a plurality of vehicles are photographed, there is a problem that only the vehicle lower end of one vehicle can be detected as in the vehicle head detection device of the first conventional example.

Further, since the apparatus of Conventional Example 2 uses only the horizontal edge for detecting the lowermost portion of the front vehicle, when many horizontal edges exist not only in the vehicle area but also on the road surface, the vehicle area There is a problem in that it is not possible to distinguish between the road surface area and the road surface area, and the vehicle lower end cannot be accurately detected. For example, when lighting is installed at regular intervals, such as in a tunnel, and there is uneven brightness on the road surface, shooting is performed obliquely to the road surface, so there is a vertical change in brightness on the screen. Since it is emphasized and the horizontal edge is detected, it is difficult to accurately detect the bottom of the vehicle.

The present invention has been made in view of the above problems, and it is possible to stably and accurately determine the upper and lower ends of a vehicle from a video captured by a camera even when there are a plurality of traveling vehicles. An object of the present invention is to provide a vehicle edge detection device that can detect the vehicle edge.

[0010]

A vehicle end detection apparatus according to the present invention includes an image input means for inputting a vehicle image, and a horizontal differential image generated from the vehicle image input by the image input means. A horizontal direction differentiating means for detecting a direction brightness change amount, a vertical direction differentiating means for detecting a vertical direction brightness change amount by generating a vertical direction differential image from the vehicle image input by the image input means, and the vertical direction Horizontal edge extracting means for extracting a horizontal edge from a differential image; edge upper image projecting means for obtaining a horizontal luminance change information by generating an edge upper projected image in an upper adjacent region of the horizontal edge from the horizontal differential image; An edge for obtaining horizontal luminance change information by generating an edge lower projection image in a lower proximity region of the horizontal edge from the horizontal differential image Partial image projection means, image difference means for calculating the difference between the edge lower projection image and the edge upper projection image to generate a difference image, and the horizontal edge is the vehicle upper end based on the information of the difference image. A horizontal edge type determining means for determining whether the vehicle is at the lower edge or neither, and an output means for outputting the vehicle upper edge and the vehicle lower edge determined by the horizontal edge type determining means. Is characterized by.

Further, the horizontal edge takes the absolute value of the pixel value for each pixel of the vertical direction differential image, and the absolute value of the pixel value is larger than the absolute values of the pixel values of other pixels in the vertical direction, A binary horizontal edge image in which the pixel value is set to the first value or the second value can be obtained depending on whether the pixel value exceeds a predetermined threshold value.

Further, the other pixels in the vertical direction may be two pixels above and below in the vertical direction.

Further, the edge upper projection image and the edge lower projection image are weighted and added to the pixel values of the horizontal direction differential image in the upper proximity region of the horizontal edge and the lower proximity region of the horizontal edge, respectively. Can be generated by

Further, the weight is a vertical distance between a pixel whose pixel value of the binary horizontal edge image is the first value and a pixel of the horizontal differential image, and d is a predetermined distance. If v, it can be a function w (d) such that w (d) = 0 when d> v.

Further, the horizontal edge type determining means sets the horizontal edge at which the change in horizontal luminance of the upper portion of the difference image is larger than the change in horizontal luminance of the lower portion as the lower edge of the vehicle,
The horizontal edge in which the lower horizontal luminance change of the difference image is larger than the upper horizontal luminance change may be determined as the upper end of the vehicle.

In the present invention, since the upper and lower ends of the vehicle are detected in the image of the vehicle based on the information of the amount of change in the brightness of the adjacent regions above and below the horizontal edge that is a candidate for the upper or lower end of the vehicle, a plurality of images are included in the image. Even if there is a vehicle, it is possible to accurately detect the vehicle end,
Since the upper and lower ends of the vehicle are detected by horizontal brightness changes that are difficult to detect in the road surface area, even if horizontal edges are detected on the road surface due to uneven brightness on the road surface, the vehicle edge and the road surface can be detected. Horizontal edges can be distinguished.

[0017]

BEST MODE FOR CARRYING OUT THE INVENTION A vehicle end detecting device according to an embodiment of the present invention will be specifically described below with reference to the accompanying drawings. FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention.

As shown in FIG. 1, in the vehicle end detection apparatus 100 of this embodiment, the image data obtained from the image input means 101 is input to the horizontal direction differentiating means 103 and the vertical direction differentiating means 104, and the horizontal direction is detected. The data processed by the differentiating means 103 and the vertical direction differentiating means 104,
Further, it is input to the upper edge image projection means 106 and the lower edge image projection means 107 together with the data processed by the horizontal edge extraction means 105. Then, the data processed by the upper edge image projecting means 106 and the lower edge image projecting means 107 are input to the image difference means 108, further input to the horizontal edge type determination means 109 and sequentially processed, and then output means 110. Is a device for detecting the upper and lower ends of a vehicle from a video image of the vehicle by being output by. These processing means for inputting data and processing it have storage means for storing the processed data respectively.

The image input means 101 is, for example, a camera installed on the roadside or above the road surface, and photographs the road surface on which the vehicle is traveling and inputs the image. The image input unit 101 is connected to the input image storage unit 111, and the input image storage unit 111 stores the vehicle image input by the image input unit 101. Further, the input image storage means 111 is connected to the horizontal direction differentiating means 103 and the vertical direction differentiating means 104, and inputs the vehicle image stored in the input image storage means 111 to the horizontal direction differentiating means 103 and the vertical differentiating means 104.

The horizontal direction differentiating means 103 and the vertical direction differentiating means 104 are connected to the horizontal direction differential image storing means 112 and the vertical direction differential image means 113, respectively, and the horizontal differentiating means 103 stores the input vehicle image. Horizontal differential processing is performed, and the horizontal differential image thus obtained is input to the horizontal differential image storage means 112. Similarly, the vertical direction differentiating means 104 performs a vertical direction differentiating process on the input vehicle image, and inputs the vertical direction differential image thus obtained into the vertical direction differential image storing means 113.

Further, the vertical differential image storage means 113
Is connected to the horizontal edge extraction means 105, and the horizontal edge extraction means 105 is connected to the horizontal edge image storage means 114. Then, the vertical differential image storage means 11
3 inputs the stored vertical direction differential image to the horizontal edge extraction means 105, and the horizontal edge extraction means 105 checks the pixel value of each pixel of the vertical direction differential image. Then, a horizontal edge image having a binary pixel value is created depending on whether or not the absolute value of the pixel value is maximum and the pixel value exceeds a predetermined threshold value. For example, a binary horizontal edge image having a pixel value of 1 when the threshold value is exceeded and a pixel value of 0 when the threshold value is less than the threshold value is created, and a horizontal edge image having a binary pixel value is generated. Horizontal edge image storage means 114
To enter.

The horizontal differential image storage means 112 and the horizontal edge image storage means 114 are both connected to the upper edge image projection means 106 and the lower edge image projection means 107, and further, the upper edge image projection means 106 and the lower edge image projection means. The means 107 is connected to the upper edge projected image storage means 115 and the lower edge projected image storage means 116, respectively. Then, the horizontal differential image storage means 112
The horizontal edge image storage unit 114 and the horizontal differential image and the horizontal edge image respectively stored therein are both the upper edge image projection unit 106 and the lower edge image projection unit 1.
Enter in 07.

The edge upper image projecting means 106 is above the horizontal edge pixels with respect to all the pixels having a pixel value of 1 in the horizontal edge image having binary pixel values (hereinafter, referred to as edge pixels). The pixel values of the pixels in the horizontal direction differential image are weighted and added, and the edge upper projection image thus obtained is input to the edge upper projection image storage means 115.

Similarly, the lower edge image projection means 107.
For all the horizontal edge pixels having a pixel value of 1 in the binary horizontal edge image, the pixel values of the pixels of the horizontal differential image below the horizontal edge pixels are weighted and added, whereby The obtained edge lower projection image is input to the edge lower projection image storage means 116.

The upper edge projected image storage means 115 and the lower edge projected image storage means 116 are both connected to the image difference means 108, connected to the horizontal edge type determination means 109 via the difference image storage means 117, and further horizontal. The edge type determination means 109 is connected to the output means 110. The edge upper projection image storage unit 115 and the edge lower projection image storage unit 116 input the stored edge upper projection image and edge lower projection image to the image difference unit 108, respectively. The image difference means 108 creates a difference image between the edge upper projection image and the edge lower projection image, and inputs it to the difference image storage means 117. Then, the difference image storage unit 117 inputs the stored difference image to the horizontal edge type determination unit 109, and the horizontal edge type determination unit 109 selects and outputs the top edge and the bottom edge of the vehicle from the difference image. Input to the means 110. Then, the output means 110 outputs the positions of the upper end and the lower end of the extracted vehicle.

The operation of this embodiment will be described in detail below. The input image obtained by the image input unit 101 and stored in the input image storage unit 111 is I (i, j).
Indicate. The coordinate system of the input image is set such that the upper left point of the image is the origin and the right direction is the positive direction of the x axis and the downward direction is the positive direction of the y axis. When the x-coordinate is i 1 and the y-coordinate is j, I (i, j) represents the luminance value of the pixel having the coordinate (i, j). FIG. 5 is a diagram showing an input image of a coordinate system in which the upper left point of the image is the origin, the right direction is the positive direction of the x axis, and the downward direction is the positive direction of the y axis. In the present embodiment, as shown in FIG. 5, a case will be described in which the rear portion of one vehicle 2 is present in the central portion of the screen 1. Vehicle 2 has window 2
a and the license plate 2b. Also, the vehicle 2
The vehicle 2 has a shadow 3 from below to the side.

FIG. 2A is a Sobel filter for horizontal differential processing, and FIG. 2B is a Sobel filter for vertical differential processing. In the present embodiment, the differentiating process detects the amount of change in luminance in the horizontal direction and the amount of change in luminance in the vertical direction.
The Sobel filter of 3 × 3 pixels shown in is used as an edge detection filter. And the horizontal differentiating means 1
The differentiation processing by 03 is performed by adding S of FIG. 2A to the digitized input image of each point of the actually obtained 3 × 3 square area.
An arithmetic process is performed in which the pixel value is multiplied by the pixel filter and integrated. The obtained horizontal differential image is Ix
Let (i, j). FIG. 6 is a diagram showing a horizontal differential image when FIG. 5 is used as an input image. A horizontal differential image is obtained as shown in FIG. 6 by performing arithmetic processing by the horizontal differential means 103 using the filter of FIG. The horizontal differential image has a small horizontal brightness change in the area other than the vehicle area on the screen 1, and the image is not displayed. The filter for differentiation is shown in FIG.
The size is not limited to 3 × 3 as in (a), but may be different sizes. Further, it is not limited to the Sobel filter as long as it can extract a luminance change in the horizontal direction.

The differentiating process by the vertical differentiating means 104 is applied to the digitized input image by, for example, S in FIG. 2 (b).
An arithmetic process is performed in which the pixel value is multiplied by the pixel filter and integrated. The obtained vertical differential image is Iy (i,
j). FIG. 7 is a diagram showing a vertical differential image when FIG. 5 is used as an input image. Similar to the horizontal direction differentiating process, the vertical direction differentiating image is obtained on the screen 1 as shown in FIG. 7 by the differentiating process of the vertical direction differentiating means 104. An image is displayed on the screen 1 in the vehicle area and at the end of the shadow of the vehicle due to a change in the luminance in the vertical direction. In addition,
The filter for differentiation is not limited to 3 × 3 as shown in FIG. 2 and may have different sizes, as in the case of horizontal differentiation. Also, not limited to the Sobel filter,
Anything can be used as long as it can obtain the luminance change in the vertical direction.

The horizontal differential image Ix (i, j) and the vertical differential image Iy (i, j) obtained by the horizontal differentiating means 103 and the vertical differentiating means 104 are respectively stored in the horizontal differential image storing means. 112 and the vertical differential image storage means 113.

The horizontal edge extraction means 105 takes the absolute value of the pixel value of each pixel of the vertical differential image Iy (i, j) stored in the vertical differential image storage means 113,
Horizontal edges are extracted by selecting the maximum of the absolute value image. The maximum, that is, the horizontal edge is selected, for example, from a pixel selected to determine whether it is the maximum (hereinafter referred to as a target pixel) and two pixels above and below the target pixel in the vertical direction, out of a total of 5 pixels. When the absolute value of the pixel value of the pixel of interest is the largest value and exceeds the predetermined threshold value, the pixel of interest is determined as the maximum and selected. When selecting the maximum value, the selection range may be changed such that the maximum value is selected from the target pixel and the upper and lower three pixels in the vertical direction from a total of seven pixels. Pixels having a maximum value, that is, a pixel determined to have a horizontal edge have a pixel value of 1, and other pixels have a pixel value of 0
Create a horizontal edge image consisting of a binary image such as FIG. 8 is a diagram showing a horizontal edge image created by extracting horizontal edges from the vertical differential image shown in FIG. 7.
On the screen 1, by the processing of the horizontal edge extraction means 105,
An image whose line width is narrower than that of the vertically differentiated image is displayed. This horizontal edge image is stored by the horizontal edge image storage means 114.

In the upper edge image projection means 106, with respect to each edge pixel having a pixel value of 1 out of the binary pixel values of the horizontal edge image, a horizontal region in a region near each horizontal edge pixel and in an upper region is provided. Directional differential image storage means 112
The projection value of the horizontal direction differential image Ix (i, j) stored in is calculated. Specifically, the weight w (d) corresponding to the vertical distance d between the edge pixel of the horizontal edge image and the pixel of the horizontal differential image Ix (i, j) is calculated as the horizontal differential image Ix (i, j). ) The calculation processing of multiplying and adding the absolute value of the pixel value is performed.

3 and 4 are graphs showing the weight w (d) with the distance d on the horizontal axis and the weight w (d) on the vertical axis. As the weight w (d), for example, as shown in the graph of FIG. 3, a weight inversely proportional to the distance d can be used.
Further, the weight w (d) may have a functional form as shown in the graph of FIG. However, the weight w (d) is set such that the distance w> d is greater than the predetermined distance v so that the pixel value of the horizontal differential image Ix (i, j) far away from the pixel of interest is not affected. Then, it is determined that w (d) = 0.

FIG. 9A is a view showing the horizontal differential image Ix (i, j) shown in FIG. 6 and the horizontal edge image shown in FIG. 8 in an overlapping manner, and FIG. 9B is a horizontal differential image Ix (i. i, j)
9A and FIG. 9B are diagrams showing the horizontal edge pixels of the area A surrounded by the ellipse in FIG. 9A in an overlapping manner, and FIGS. 9C and 9D respectively show an edge upper projection image and an edge lower projection image of the area A, respectively. It is a graph which shows an image.

In FIG. 9A, when the edge upper image projecting means 106 processes the horizontal edge pixels of the area A near the vehicle center surrounded by the ellipse of the screen 1, the horizontal edge pixels of the area A are selected. On the other hand, the weight w (d) becomes a value other than 0 in the region B surrounded by the rectangle existing above the center of the vehicle as shown in FIG. 9B. At this time, the height of this rectangle is a predetermined distance v. A calculation process is performed on the absolute value of the horizontal differential image Ix (i, j) of the rectangular area B using the weight w (d).

In this calculation process, the coordinates of the horizontal edge pixel are (i, j) and the edge upper projection image is Iaproj (i, j).
Then, Iaproj (i, j) is obtained by the following formula 1. It should be noted that, in a pixel having a pixel value of 0 in the horizontal edge image, the edge upper projection image Iaproj (i, j) is 0.

[0036]

[Equation 1]

At the position corresponding to the horizontal edge pixel of the area A, the edge upper projection image Iaproj obtained by the equation 1 is obtained.
As shown in FIG. 9C, the projected value of (i, j) becomes larger as the luminance change is larger above the horizontal edge pixel. In FIG. 9C, the vertical axis represents the projection value and the horizontal axis represents the position in the x direction on the horizontal edge.

Similar to the upper edge image projecting means 106, the lower edge image projecting means 107, for each edge pixel, is a horizontal differential image Ix (i, i, i, i) in an area near the edge pixel and below the edge pixel. j) Obtain the projection value. A weight w is applied to the area of the horizontal differential image Ix (i, j) projected by the edge lower image projection means 107, that is, to each edge pixel of the area A, on the pixel at the position corresponding to the horizontal edge pixel of the area A. An area where (d) has a value other than 0 is a rectangular area C below the vehicle center shown in FIG. 9B. The height of this rectangle is the same predetermined distance v as the area B.

Here, the lower edge projection image is Ibproj
Assuming that (i, j), Ibproj (i, j) is the following Equation 2
Is obtained by

[0040]

[Equation 2]

The weight w (d) in the equation 2 shows the same function as w (d) in the edge upper image projection means 106. Further, Ibproj (i, j) is 0 in a pixel having a pixel value of 0 in the horizontal edge image. At the position corresponding to the horizontal edge pixel of the area A, the projection value of the lower projection image Ibproj (i, j) obtained by Equation 2 is as shown in Fig. 9 ( d) . The larger the change in brightness, the larger the value. Also in FIG. 9D, FIG.
Similar to (c), the vertical axis represents the projection value and the horizontal axis represents the position in the x direction on the horizontal edge.

Hereinafter, a group of adjacent horizontal edge pixels is referred to as an edge. Since there is little change in the brightness of the road surface area on the image, but there is a large change in the brightness of the vehicle area, if the area above the edge is the road surface area or the shadow area, the upper projection image Iaproj (i, j) The projection value takes a value close to 0. On the other hand, when the vehicle is above the edge, the projection value of the upper projection image Iaproj (i, j) is large.
Similarly, when the lower side of the edge is a road surface area or a shadow area, the projection value of the lower projection image Ibproj (i, j) takes a value close to 0. On the other hand, when there is a vehicle below the edge, the projection value of the lower projection image Ibproj (i, j) is large.

The image difference means 108 creates a difference image by subtracting the edge lower projection image from the edge upper projection image thus obtained, and the difference image storage means 117 stores it. Assuming that the difference image processed by the image difference unit 108 and stored in the difference image storage unit 117 is Isub (i, j), Isub (i, j) is obtained by Equation 3 below.

[0044]

[Equation 3]

FIG. 10 is a graph showing the difference image created by the image difference means 108. Image difference means 108
On each horizontal edge shown in FIG. 8 obtained by the processing in FIG. 8, for the projection value obtained by the projection as shown in FIG. 9, as shown in FIG. The difference value is obtained by subtracting the edge lower projection value from the value. The difference image storage means 117 stores a difference image Isub (i, j) in which the pixel value of the pixel corresponding to each horizontal edge is the difference value. If the difference image Isub (i, j) does not correspond to a horizontal edge,
The pixel value of the pixel of the difference image Isub (i, j) is 0.

FIG. 11A is a diagram showing an input image of the vehicle,
11B to 11E are graphs showing the difference values corresponding to the horizontal edges of the plurality of regions of the input image shown in FIG. 5, which are the vehicle upper end portion, the vehicle region, the vehicle lower end portion, and the vehicle upper end portion, respectively. The difference value in the vehicle side part is shown.

When there is a vehicle above the edge and a road surface (not shown) below the edge, the difference value takes a positive value, and such an edge corresponds to the lower end of the vehicle. As shown in FIG. 11D, the difference value on the edge corresponding to the vehicle lower end 4 of the vehicle 2 is a positive value. On the other hand, when the vehicle is below the edge and the road is above the edge, the difference value takes a negative value, and such an edge corresponds to the upper end of the vehicle. Figure 11
As shown in (b), the value of the difference value on the edge corresponding to the vehicle upper end 5 of the vehicle 2 is a negative value.

On the other hand, when there is a horizontal edge inside the vehicle region, there are brightness changes both above and below, so the value takes both positive and negative values depending on the location, and the entire edge becomes positive or negative. The probability is low, the positive region and the negative region are mixed, and the length of the positive region and the length of the negative region become short. As shown in FIG. 11C, the difference value on the horizontal edge inside the vehicle is such that the positive region and the negative region are mixed at a short interval.

When the horizontal edge is extracted because the upper or lower part is the shadow area, the difference value on the horizontal edge is small because the brightness change is small in both the upper and lower parts of the shadow area or the road surface. The value of is close to 0 for the entire edge. Therefore, as shown in FIG. 11E, the difference value on the horizontal edge due to the shadow is close to 0 for the entire edge.

FIG. 12 is a diagram showing a difference image created from the upper edge projected image and the lower edge projected image.
The difference value image can be displayed on the screen as, for example, a pixel having a positive difference value as white, a pixel having a negative difference value as black, a pixel having a difference value of 0 as gray, and the like. In that case, as shown in FIG. 12, a difference image in which the horizontal edge is color-coded into the white region 6 or the black region 7 is obtained on the screen 1. The area other than the horizontal edge is the gray area 8. The horizontal edge of the vehicle upper end 5 shown in FIG.
The horizontal edge of the vehicle area shown in (c) is shown as an area in which the white area 6 and the black area 7 are mixed. In addition, the horizontal edge of the vehicle lower end 4 shown in FIG.
The horizontal edge due to the shadow shown in 1 (e) has a difference value of almost 0, and thus becomes the gray area 8.

The difference image thus generated by the image difference means 108 is stored in the difference image storage means 117 and input to the horizontal edge type determination means.

In the horizontal edge type determining means 109, first, only the edge pixels having a positive difference value on the horizontal edge and exceeding the preset threshold value Ith of the brightness value are selected from the difference image to be adjacent to each other. An edge that is a group of edge pixels to be obtained is obtained. Then, among the obtained edges, only those whose edge length exceeds a preset threshold value are selected and set as vehicle lower end candidates. If the edge length does not exceed the threshold value, it is erased. The length of the edge is defined by the difference between the maximum value and the minimum value of the coordinate value x of the pixels forming the same edge. Then, the vehicle aspect ratio r = (vehicle height) /
(Vehicle width) is set in advance, and for each vehicle lower end candidate, the upper side of the vehicle of the vehicle lower end candidate, that is, FIG.
It is checked whether or not there is another vehicle lower end candidate within the region of the length xr of the vehicle lower end candidate in the negative direction of the y direction defined in (4). At this time, if another vehicle lower end candidate exists above the vehicle lower end candidate, the lengths of both edges are compared, and if the upper lower end candidate has a shorter length, the upper lower end candidate is deleted. When the lengths of the upper and lower edge candidates are long or similar, the average of the difference values of both edges is calculated, and (the average value of the difference values of the lower edges) / (the difference value of the upper edge is calculated). If the average value) is larger than a preset threshold value, the upper edge is erased, and if it is smaller, the lower edge is erased. Then, the remaining edge is detected as the lower end of the vehicle. The upper end of the vehicle is detected by performing similar processing on an edge whose difference value on the edge is negative and the length of the edge exceeds the threshold value.

In this way, the horizontal edge type determining means 109 detects the edges of the vehicle upper end and the vehicle lower end. Then, the vehicle upper end and the vehicle lower end, which are determined and detected by the horizontal edge type determination means 109, are input to the output means 110.

The output means 110 outputs the detected position of the vehicle end, and is, for example, the input image storage means 111.
The detected vehicle edge is superimposed on the input image stored in, and displayed on the monitor. Alternatively, the position of the vehicle end detected on the console may be displayed numerically.

According to the present embodiment, since the change in the brightness of the road surface area on the image is small, but the change in the brightness of the vehicle area is large, the upper side of the edge is the road surface area or the shadow area. If there is, the upper edge projection image Iaproj
While (i, j) takes a value close to 0, when there is a vehicle above the edge, the edge upper projection image Iaproj (i, j
The value of j) becomes large, and when the vehicle is below the edge, the value of the edge lower projection image Ibproj (i, j) becomes large. Therefore, the vehicle upper end and lower end can be detected from the value of the difference image Isubj (i, j) obtained from the difference value between the edge upper projection image and the edge lower projection image. That is, if it is the upper end of the vehicle, the upper edge projection image Iaproj
(I, j) is small and the edge bottom projection image Ibproj (i,
j) is large, the difference value is negative, and if the vehicle is a lower part, the edge upper projection image Iaproj (i, j) is large and the edge lower projection image Ibproj (i, j) is small, so the difference value is a positive value. Therefore, the vehicle upper end can be detected by selecting the edge having the negative difference value, and the vehicle lower end can be detected by selecting the edge having the positive difference value.

[0056]

As described in detail above, according to the present invention,
The upper and lower ends of the vehicle are detected by utilizing the fact that there are few horizontal brightness changes on the road surface, but many horizontal brightness changes exist in the vehicle region. That is, the amount of horizontal luminance change existing in a limited area near the upper and lower edges of the horizontal edge which may be the upper and lower edges of the vehicle is examined, and the horizontal luminance change is large in the upper part but small in the lower part. An edge is detected as the lower end of the vehicle, and an edge with a large horizontal brightness change in the lower part but little in the upper part is detected as the upper end of the vehicle. As a result, the accuracy of the upper and lower ends of each vehicle is stable even when there are multiple vehicles in the image, without being affected by changes in brightness caused by objects other than the vehicles located far from the upper and lower ends of the vehicle. You can ask well.

[Brief description of drawings]

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

FIG. 2A is a Sobel for horizontal direction differential processing.
Filter, (b) Sobe for vertical differential processing
It is a figure which shows a l filter.

FIG. 3 shows the distance d on the horizontal axis and the weight w (d) on the vertical axis,
It is a graph which shows weight w (d).

FIG. 4 is a graph diagram showing the weight w (d) with the distance d on the horizontal axis and the weight w (d) on the vertical axis.

FIG. 5 is a diagram showing an input image of a coordinate system in which the upper left point of the image is the origin, the right direction is the positive direction of the x axis, and the downward direction is the positive direction of the y axis.

FIG. 6 is a diagram showing a horizontal differential image of an input image.

FIG. 7 is a diagram showing a vertical differential image of an input image.

FIG. 8 is a diagram showing a horizontal edge image created by extracting horizontal edges from a vertical differential image.

9A is a diagram in which a horizontal differential image Ix (i, j) and a horizontal edge image are overlapped with each other, and FIG. 9B is a horizontal differential image Ix (i, j) and a horizontal edge of an area A. The figure which overlapped with the pixel and (c) and (d) is a graph figure which shows the edge upper projection image and edge lower projection image of the area | region A, respectively.

FIG. 10 is a graph showing a difference image created by the image difference means.

11A is a diagram showing an input image of a vehicle, and FIGS. 11B to 11E are graphs showing difference values corresponding to horizontal edges of a plurality of regions in the vehicle region of the input image. , The difference values at the vehicle upper end portion, the vehicle region, the vehicle lower end portion, and the vehicle side portion, respectively.

FIG. 12 is a diagram showing pixel values of a difference image between an upper edge projected image and a lower edge projected image.

[Explanation of symbols]

1; screen 2; vehicle 3; Shadow 4; Bottom of vehicle 5; Vehicle top 6; white area 7: Black area 8: Gray area 101; Image input means 103; horizontal differentiating means 104: Vertical direction differentiating means 105; Horizontal edge extraction means 106; Edge upper image projection means 107; bottom edge image projection means 108; Image difference means 109: Horizontal edge type determination means 110; Output means 111; Input image storage means 112; Horizontal differential image storage means 113; Vertical direction differential image storage means 114; Horizontal edge image storage means 115; Edge upper projection image storage means 116; Bottom edge projection image storage means 117; Difference image storage means

─────────────────────────────────────────────────── --- Continuation of the front page (56) References JP-A-8-43083 (JP, A) JP-A-6-331335 (JP, A) JP-A-9-198512 (JP, A) JP-A-7- 182484 (JP, A) JP 5-282593 (JP, A) JP 7-28975 (JP, A) JP 2001-134772 (JP, A) JP 2000-113201 (JP, A) JP 2000-136905 (JP, A) JP 2000-266539 (JP, A) JP 2001-5979 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) G06T 1/00 330 G08G 1/00 JISST file (JOIS)

Claims (6)

(57) [Claims] 1. An image inputting means for inputting a vehicle image; a horizontal differentiating means for generating a horizontal differential image from the vehicle image input by the image inputting means to detect a horizontal luminance change amount; Vertical differential means for generating a vertical differential image from the vehicle image input by the image input means to detect a vertical luminance change amount; horizontal edge extracting means for extracting a horizontal edge from the vertical differential image; Edge top image projection means for obtaining horizontal luminance change information by generating an edge top projection image in the upper proximity region of the horizontal edge from the horizontal direction differential image,
An edge lower image projecting unit for obtaining horizontal luminance change information by generating an edge lower projected image in a lower proximity region of the horizontal edge from the horizontal direction differential image; and the edge lower projected image and the edge upper projected image. An image difference unit that calculates a difference and generates a difference image, and a horizontal edge type that determines whether the horizontal edge is the vehicle upper end, the vehicle lower end, or neither based on the information of the difference image A vehicle end detection device comprising: a determination unit; and an output unit that outputs the vehicle upper end and the vehicle lower end determined by the horizontal edge type determination unit. 2. The horizontal edge takes an absolute value of a pixel value for each pixel of the vertical direction differential image, and the absolute value of the pixel value is larger than the absolute values of pixel values of other pixels in the vertical direction, The vehicle end according to claim 1, wherein the pixel value is a binary horizontal edge image in which the pixel value is set to a first value or a second value depending on whether or not a predetermined threshold value is exceeded. Detection device. 3. The vehicle end detection device according to claim 2, wherein the other pixels in the vertical direction are two pixels in the vertical direction and two pixels in the vertical direction. 4. The edge upper projection image and the edge lower projection image are weighted and added to the pixel values of the horizontal differential image in the upper proximity area of the horizontal edge and the lower proximity area of the horizontal edge, respectively. The vehicle end detection device according to claim 1, wherein the vehicle end detection device is generated by the above method. 5. The horizontal edge is the vertical differential image.
For each pixel of the image, take the absolute value of the pixel value
The absolute value of is the absolute value of the pixel value of other pixels in the vertical direction .
Depending on whether or not it exceeds the specified threshold
Binary water whose pixel value is set to the first value or the second value
The weight is a flat edge image, and the weight is a distance in the vertical direction between a pixel in which the pixel value of the binary horizontal edge image is the first value and a pixel in the horizontal direction differential image, and a predetermined distance. Is a function w (d) such that w (d) = 0 when d> v, the vehicle end detection device according to claim 4 . 6. The horizontal edge type determining means sets the horizontal edge of the upper portion of the difference image whose horizontal luminance change is larger than the lower horizontal luminance change as a lower edge of the vehicle, and sets the lower horizontal luminance of the difference image. 6. The vehicle end detection device according to claim 1, wherein the horizontal edge whose change is larger than the change in horizontal luminance of the upper part is determined as the upper end of the vehicle.