JPH09128548A - Vehicle detecting method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To exactly and easily recognize a vehicle based on a binarized image, to by detecting the centroid position of a candidate vehicle area and excluding this area from the recognizing object of the vehicle when the centroid position is out of a prescribed range. SOLUTION: A camera 1 outputs an image signal corresponding to a picked-up monochromatic image to an image processor 3. In the image processor 3, internal holes or neighbor black areas are collected into one lot and binarized by expanding black sections in the image and afterwards, noise removing processing is executed for removing mixed areas. Then, the centroid position of the black area is found and when the centroid position is out of a prescribed range, that black area is excluded from the recognizing object as the vehicle. When the centroid position is inside the prescribed range, preferably further, parameters such as the aspect ratio of the black area, area ratio between the real area of the black area and the area of a rectangular area in contact with the black area and the height of highest part in the black area are successively found and when any one of these parameters is correspondent to prescribed conditions, that black area is excluded from the recognizing object as the vehicle.

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention relates to a vehicle detection method.

[0002]

2. Description of the Related Art In recent years, a control method for recognizing an obstacle or a vehicle in front of the vehicle and avoiding the obstacle or optimizing an inter-vehicle distance to allow safe traveling when the vehicle is traveling has been proposed. It is coming to be proposed.

For example, as a method of recognizing a vehicle ahead,
There is a method of extracting a vehicle by utilizing the characteristic of the horizontal component of the rear view of the vehicle in front. Another method for extracting obstacles and vehicles is to use a camera provided in front of the vehicle to take a front view and use edge information of an object in the image obtained by the camera, or There is a method of using distance information to the object.

As a method of utilizing edge information, a region (vertical edge information) sandwiched between vertical line regions in an image is set as a vehicle candidate, and the information satisfies conditions such as vehicle size and left-right symmetry. There is a method of recognizing it as a vehicle when it is doing. Further, a method of recognizing a region (horizontal edge information) including many horizontal line segments in a captured image as a vehicle has been proposed (for example, Japanese Patent Laid-Open No. 3-273500).

Further, after distance measurement is performed on each of a large number of areas of the photographed image to obtain distance information, if there is an area having the same distance in the adjacent areas in the distance information, the area is detected. There is a method of recognizing as an obstacle or an object such as a vehicle.

[0006]

However, in the above-described method for recognizing a vehicle or an obstacle, for example, when using edge information, in the case of a short distance, an obstacle or a vehicle is detected in the image. Obstacles and vehicles are relatively easy to recognize because they are easy to draw clearly, but it becomes difficult to detect edges because the difference in contrast between obstacles in front of the vehicle or the vehicle and the surroundings decreases as the distance increases. This makes recognition as a vehicle or an obstacle difficult or inaccurate.

In particular, when a photographed image obtained as a front vehicle is used as a gradation image and the gradation image is converted into a binary image by setting a threshold value, a side fence or the like is provided near the front vehicle. If the shadow from is present, an image including the shadow will be obtained. Therefore, the vehicle area at the time of recognizing the vehicle has a size including a portion other than the vehicle such as the above-described shade as shown by a chain double-dashed line in FIG. 16, and the vehicle ahead is actually present. However, the size of the vehicle area may cause a situation in which the vehicle area is excluded from the recognition target as a vehicle, and the vehicle detection based on the vehicle area tends to be inaccurate.

Therefore, in view of the problems in the conventional vehicle detection method described above, an object of the present invention is to provide a vehicle detection method capable of accurately and easily recognizing a vehicle based on a binarized image. To do.

[0009]

To achieve the above object, the invention according to claim 1 binarizes a road image photographed by a camera, and extracts a vehicle candidate region from a black region of the binarized image. Then, in the vehicle detection method for recognizing the vehicle on the road based on the vehicle candidate area, the center of gravity position of the vehicle candidate area is detected, and when the same center of gravity position is outside the predetermined range, the vehicle candidate area The feature is that it is excluded from the recognition target.

According to a second aspect of the present invention, a road image captured by a camera is binarized, a vehicle candidate area is extracted from a black area of the binarized image, and the vehicle on the road is based on the vehicle candidate area. The vehicle detection method for recognizing is characterized by detecting the aspect ratio of the vehicle candidate area, and excluding the vehicle candidate area from the recognition target of the vehicle when the aspect ratio is equal to or greater than a set value.

According to a third aspect of the present invention, a road image captured by a camera is binarized, a vehicle candidate area is extracted from a black area of the binarized image, and the vehicle on the road is based on the vehicle candidate area. In the vehicle detection method for recognizing, the actual area of the vehicle candidate area and the area of the rectangular area in contact with the vehicle candidate are detected, and when the area ratio of the actual area to the area of the rectangular area is equal to or less than a set value, It is characterized in that the vehicle candidate area is excluded from the recognition target of the vehicle.

According to a fourth aspect of the present invention, a road image captured by a camera is binarized, a vehicle candidate area is extracted from a black area of the binarized image, and the vehicle on the road is based on the vehicle candidate area. In the vehicle detection method for recognizing, the uppermost coordinate height of the vehicle candidate area is detected, and when the uppermost coordinate height is equal to or more than a set value, the vehicle candidate area is excluded from the recognition target of the vehicle. Is characterized by.

According to a fifth aspect of the present invention, a road image captured by a camera is binarized, a vehicle candidate area is extracted from a black area of the binarized image, and the vehicle on the road is based on the vehicle candidate area. In the vehicle detection method for recognizing the above, the actual area of the vehicle candidate area is detected, and when the actual area is outside the set range, the vehicle candidate area is excluded from the recognition target of the vehicle.

According to a sixth aspect of the present invention, a road image captured by a camera is binarized, a vehicle candidate area is extracted from a black area of the binarized image, and the vehicle on the road is based on the vehicle candidate area. In the vehicle detection method for recognizing the above, the horizontal length of the vehicle candidate area is detected, and when the horizontal length is outside the setting range, the vehicle candidate area is excluded from the recognition target of the vehicle. .

According to a seventh aspect of the present invention, in the vehicle detection method according to the sixth aspect, the upper limit value and the lower limit value of the setting range are set to be smaller as the image height of the vehicle candidate region is higher. It has a feature.

According to an eighth aspect of the present invention, a road image captured by a camera is binarized, a vehicle candidate area is extracted from a black area of the binarized image, and the vehicle on the road is based on the vehicle candidate area. In the vehicle detection method for recognizing, the horizontal position center position of the vehicle candidate region and the center of gravity position of the vehicle candidate region are detected, and when the horizontal deviation between the center position and the center of gravity position is a predetermined value or more. Is characterized by excluding the vehicle candidate region from the recognition target of the vehicle.

According to a ninth aspect of the present invention, a road image captured by a camera is binarized, a vehicle candidate area is extracted from a black area of the binarized image, and the vehicle on the road is based on the vehicle candidate area. In the vehicle detection method for recognizing the above, the inertial principal axis of the vehicle candidate area is detected, and when the inclination of the inertial principal axis is a predetermined value or more, the vehicle candidate area is excluded from the recognition target of the vehicle.

According to a tenth aspect of the present invention, a road image captured by a camera is binarized, a vehicle candidate area is extracted from a black area of the binarized image, and the vehicle on the road is based on the vehicle candidate area. In the vehicle detection method for recognizing, the detection of the center of gravity of the vehicle candidate area, the detection of the aspect ratio in the vehicle candidate area, the detection of the area ratio between the actual area of the vehicle candidate area and the area of the rectangular area in contact with the vehicle candidate area. Detecting the top coordinate height of the vehicle candidate area, detecting the actual area of the vehicle candidate area, detecting the horizontal length of the vehicle candidate area, the center position of the vehicle candidate area in the horizontal length direction, and the center of gravity position of the candidate area. Horizontal deviation of the vehicle and the inertial axis of the vehicle candidate area are continuously detected, and when each detection result meets the vehicle candidate exclusion condition set for these detection results, the vehicle candidate area is detected. The above car It is characterized by excluding from the recognition target.

[0019]

The present invention will be described in detail below with reference to illustrated embodiments.

FIG. 1 is a diagram showing a system configuration used in the vehicle detection method according to the present invention, and FIG. 2 is a flow chart for explaining a process used in the vehicle detection method according to the present invention.

In FIG. 1, reference numeral 1 is a camera composed of a CCD sensor, 3 is an image processing device for processing an image of the camera 1 into data that can be calculated or displayed, and 5 is a traveling path. An arithmetic unit composed of a computer capable of obtaining required measurement data such as the position of an object in 6), 6 is an input / output storage device (RAM) for storing various data, and 7 is an output for storing an arithmetic program. A dedicated storage device (ROM), 2 is a display device including a display for displaying a captured image and calculation or measurement results, and 4 is a system bus for transmitting data between the above elements. The camera 1 is installed at a predetermined location of the host vehicle, for example, in the vicinity of the rearview mirror, so as to capture the front of the host vehicle, and can capture monochrome image information.

The vehicle detection method with the system configuration shown in FIG. 1 is executed based on the processing procedure shown in FIG.

In FIG. 2, first, based on the image data from the camera 1, the black part in the photographed monochrome image is expanded (S1), and the expanded image is binarized to obtain a black area and a white area. It is divided into areas (S2), noise removal processing is further performed (S3), the vehicle-likeness of the black area after noise removal is checked, and the portion not corresponding to the vehicle recognition target is excluded (S4).

Of the processing in each of the above steps, the method described in the specification of Japanese Patent Application No. 7-166399 filed by the applicant of the present application is used for the processing in steps S1 to S3.

The contents will be described below. FIG. 3 is a view in front of the vehicle as seen from the driver's seat while the vehicle is traveling. The camera 1 images a landscape similar to the landscape shown in FIG. The camera 1 outputs an image signal corresponding to the captured monochrome image to the image processing device 3. In addition,
In FIG. 3, the dark portion is shaded and the darker portion is displayed with a finer pitch. Also, the lower part of the front vehicle is dark due to the shadow, and the tire is black, so it is much darker than the other parts, and it is represented by being painted black.

In the image processing device 3, the black portion (including the gray portion) in the image shown in FIG. 3 is expanded, and as shown in FIG. 4, a hole (white portion) inside the black portion and a neighboring portion are formed. The black areas are combined into one area.
In the expansion process in this case, the background pixels are converted into graphic pixels,
This is a method for gathering holes (white portions) included in the black portion and neighboring black areas into one area. In the case of the present embodiment, an area (matrix) in which a binary image can be captured in the vertical and horizontal directions is provided. The 8-neighbor expansion processing, which is set for every three pieces, is used. As is well known, the 8-neighborhood processing is performed when there is at least one graphic pixel 1 (a black portion by binarization) in the 8-neighborhoods P _{1 to} P ₈ of a certain background pixel (P ₀ ).
₀ gradation value f a (P ₀₎ to "1" refers to the process of converting the background pixels in the graphic pixel (e.g., issued Sep. 30, 1994, Shokodo published Inc., "Image Processing and Recognition ”, pp. 70-71).

In the present embodiment, as shown in FIG. 5, in a region (A to I) composed of 9 pixels, E = MIN (A, B, C, D, F, G, H, I) The processing of (corresponding to the logical sum in the logical operation) is sequentially performed to shrink the bright portion (area).

That is, in FIG. 5, among the nine pixels A to I, the central pixel (E) is focused, and the pixels A to D and FI around the pixel (E) are surrounded. If there is a pixel having a luminance value smaller (lower) than the luminance value of the E pixel, the E pixel is also adjusted to the luminance value of the pixel having the smallest luminance value. Since the pixel E is also included in the above equation, when the luminance value of the pixel of E is the smallest, the luminance value does not change. Scanning is performed by moving the pixels that come to the center of the matrix one by one, and the brightness value of the pixels at the center is sequentially set.

In the case of the present embodiment, when the monochrome image shown in FIG. 3 is composed of 512 × 448 pixels, all these pixels are scanned. As a result of such scanning, as shown in FIG. 4, the size of the black region becomes larger than the initial size and is expanded.

After the expansion process to obtain the image shown in FIG.
By binarizing, an image of a white area and a black area as shown in FIG. 6 is obtained. In the image divided into the black and white area, a portion that should not be originally depicted on the image as a vehicle, for example, holes mixed in the black area and vehicles other than vehicles located near the black area corresponding to the vehicle to be detected There are cases where black areas are mixed. Therefore, noise removal processing is performed to remove these mixed areas.

The noise removal process is a process for updating the tone value of the target pixel on the image to a simple average value of the tone values in the vicinity thereof to smooth the change of the tone value in the image.

Specifically, as disclosed in the above specification, the black region after binarization has a radius r1 (for example, 5).
After expanding with a circle of radius of pixel size), it is further reduced with the radius to restore the original size. Now, such expansion / contraction processing is referred to as "processing 1". The expansion of the black area is
Image-wise, the outline (outer periphery) that forms the black area
A circle with radius r1 is rolled up, and the locus drawn by the circle is set as a new black area. As a result, as described above, the holes (bright portions) inside the black area and the adjacent black areas can be integrated. Next, the expanded black area is contracted by the circle having the radius r1. Imagewisely, the contraction processing in this case rolls the circle having the radius r1 inside the contour forming the expanded black region, and forms a new black region inside the locus drawn by the circle. As a result, the black area becomes the original size. By such expansion processing, holes and the like located in the black area are gathered and connected in the black area to obtain a new black area, and further contraction processing is performed to create a black pixel corresponding to a graphic pixel in the image. By contracting the connected component of the region inward by one pixel, a new black region is obtained inside the locus of radius r1.

By the way, when a hole or a neighboring black area is connected to the inside of a black area and combined into one by the above-mentioned "Processing 1", even objects which are rarely not combined into one are combined into one. Therefore, it is necessary to fractionate these in order to restore them. Therefore, the black area after “Processing 1” is reduced with a circle having a predetermined radius r2 (for example, a radius of 3 pixels), and then expanded again with the circle having the radius r2 to have the original size. Return to. That is, when the contraction process is performed with the circle having the radius r2, the connected component of the black region is contracted inward by one pixel and removed as in the contraction process described above.
After the removal, the expansion process is performed again to restore the original size of the black region in the state where the connection portion is removed. Now, such a process is referred to as "process 2".
The “process 2” is similar in image to the “process 1” described above.

The above-mentioned "Processing 2" will be described in more detail as follows. 7 to 10 are diagrams for explaining the "process 2". The binary image shown in FIG. 7 is for explaining the noise removal process, and the binary image shown in FIG. It is an image different from the image.

In FIG. 7, four black areas 20-23
Means different objects that you do not want organized. However, when such a black region is subjected to the above-mentioned “processing 1”, as shown in FIG.
And the group of 21 and the group of 22 and 23 are the connecting portions 24, 2.
It will be connected by 5. Therefore, the black areas 20 to 2
When 3 is reduced by a circle having a radius r2 as described above, the connecting portions 24 and 25 are removed as shown in FIG.
And 21 and the black areas 22 and 23 are separated. Then, the separated black regions 20 to 23 are returned to the original size in a separated state.

The image shown in FIG. 11 corresponds to the black area after the above-described noise removal processing is performed on the binary images 26 to 30 shown in FIG. In FIG.
The vehicle-likeness in the black area is checked to exclude the objects other than the vehicle recognition target.

The parameters used for excluding objects other than the recognition target as a vehicle are as follows. (1) Position of center of gravity of black area corresponding to vehicle candidate area (2) Aspect ratio of black area (3) Area ratio between actual area of black area and area of rectangular area in contact with the black area (4) Of black area Top coordinate height (5) Real area of black area (6) Horizontal length of black area (7) Horizontal deviation between center position of black area in horizontal length direction and center of gravity of black area (8) Black The inclination of the inertial principal axis of the region These parameters are sequentially obtained, and if any of them meets the following conditions with respect to a predetermined range or a predetermined value, they are excluded from the recognition target as a vehicle. Therefore, if all of the above parameters are satisfied until the following conditions are met, the black area used as the recognition target as a vehicle is excluded from the recognition target without determining each parameter thereafter. The work is completed.

First, the barycentric position of the black area, which is the parameter mentioned in (1), is obtained, and when the barycentric position is outside the predetermined range, the black area is excluded from the recognition target as the vehicle.

Normally, the front vehicle seen from the driver's seat is shown in FIG.
As shown in, the image appears at the center of the screen or slightly downward from the center. Therefore, the position of the center of gravity in the black region representing the lower part of the vehicle ahead is located at the center or a range slightly below the center. Therefore, the black areas 26 and 27 of the black areas 26 to 30 in FIG. 11 can be determined not to be a vehicle, whereas the black areas 28 to 30 can be determined.
Can be determined to be a vehicle.

When the position of the center of gravity, which is the parameter listed in (1), is within the predetermined range, the secondary aspect ratio of the black area, which is the parameter listed in (2), is detected, and the aspect ratio is equal to or greater than the predetermined value. In that case, the black region is excluded from the recognition target as a vehicle. The size of the vehicle has an aspect ratio within a certain range, and in the case of an ordinary vehicle as shown in FIG. 3, the black area in the image after being binarized and noise-removed is as described above. Since this corresponds to the part of the tire including the lower part of the vehicle, this part is depicted in landscape. However, depending on the vehicle, the vertical and horizontal lengths may be the same or may be vertical. In such a case, a vehicle with a black hood or a cover is applicable, and the black area may be vertically long.

Therefore, in this embodiment, the aspect ratio is set to 1: 1.
The range is set to a certain range, and when it is within this range, it is recognized as a vehicle, and when it is outside this range, it is excluded from the recognition target as a vehicle.

When the aspect ratio, which is the parameter listed in (2), is within a predetermined range, the ratio of the actual area of the black region, which is the parameter listed in (3), to the area of the rectangular region with respect to the black region. Is detected, and when the area ratio is equal to or less than a preset value, the black area is excluded from the recognition target as the vehicle.

FIG. 12 is a schematic diagram showing a method for detecting the parameters mentioned in (3). In FIG.
The black area in the value image is a depiction of the lower portion of the vehicle as described above. Therefore, the ratio of the black area to the vehicle area A ′ obtained from the black area A becomes high, and the ratio is detected as the area ratio. The ratio of the black area to the vehicle area is high in FIG. 12 (A).
When the distance to the vehicle in front is shown in Fig. 12 and Fig. 12
It corresponds to any of the cases where the distance to the vehicle in front shown in (B) is long.

In this embodiment, the ratio of the area of the black area (object) corresponding to the candidate vehicle to be recognized and the area of the vehicle area obtained from the area of the black area is defined as the object area / vehicle area area ≧ The relationship of 0.5 is set. As a result, as shown in FIG. 16, it is possible to exclude the case where the vehicle area based on the black area when the shadow of the sidewall or the like is included becomes extremely large, so that the shadow exists near the vehicle. However, it is possible to prevent erroneous detection when recognizing a black area as a vehicle.

When the area ratio, which is the parameter listed in (3), has a predetermined value, the uppermost coordinate position of the black area, which is the parameter listed in (4), is detected, and the position is greater than or equal to the set value. If there is, the black area is excluded from the recognition target as the vehicle. This is based on the premise that the black region in the image after being binarized and subjected to noise removal corresponds to the portion of the tire including the lower portion of the vehicle, as described above. In such a case, in FIG. 9, the uppermost portions 20a and 20b of the black region 20 having a substantially V shape upward from the substantially center of the screen.
The coordinates of may be located above the screen and above the center right.

When the coordinate height, which is the parameter listed in (4), is less than or equal to the set value, the area of the black region, which is the parameter listed in (5), is detected and the area is outside the set range. Then, the black area is excluded from the recognition target as the vehicle. In this case, the actual area of the black region is outside the range of the area set assuming the normal inter-vehicle distance between the preceding vehicle and the host vehicle, for example, when it is small, it is determined that the vehicle is not clearly the vehicle. It is like this.

When the actual area, which is the parameter listed in (5), is within the setting range, the horizontal length of the black area, which is the parameter listed in (6), is detected and the length is outside the setting range. In case of, the black area is excluded from the recognition target as the vehicle.

In this embodiment, it is assumed that the horizontal length of the vehicle on the screen can be predicted to some extent on the coordinates.
The vehicle is not recognized when the horizontal length of the position occupied by the black area is extremely long or, on the contrary, extremely short. In particular, the upper limit value and the lower limit value of the setting range are set smaller as the height of the black region in the image increases. Thus, even if the height of the black area on the image surface changes according to the distance between the front vehicle and the host vehicle, the horizontal setting range corresponding to the height can be standardized. .

The following relational expression is used as a reference for recognizing a black area as a vehicle from the horizontal length. W _- max = f (y) (f (y) is a function determined by the position on the coordinates) W _- min = f (y) × k (coefficient of k = 0.2 to 0.3) The horizontal length of the black area (object) to be recognized as is: W _- min ≤ horizontal length of object ≤ W _- max The horizontal length which is the parameter given in (6) is within the setting range. In some cases, the horizontal center position of the black region and the barycentric position of the black region, which are the parameters listed in (7), are detected, and the horizontal deviation between the center position and the barycentric position is greater than or equal to a predetermined value. Then, the black area is excluded from the recognition target as a vehicle.

As shown in FIG. 6, in the black area to be recognized as the vehicle, the tire portion of the vehicle is depicted as the black area when the image is a binary image. Therefore, the black area has a substantially rectangular shape or a rectangular shape close to an ellipse as shown in FIG.

Therefore, as shown in FIG. 13A, when the object is recognized as a vehicle, the position of the center of gravity obtained from the shape of the black area is substantially the center position of the black area in the horizontal direction. As shown in FIG. 13B, in the case of a black area corresponding to the shadow of a sidewall other than the vehicle, the shape of the black area was obtained by mapping on the coordinates. In this case, the position of the center of gravity deviates from the center position of the black area.

In the present embodiment, the horizontal length indicated by the symbol Hw in FIG. 13B is detected, and the center of gravity position indicated by the symbol Δw in FIG. 13B is detected with respect to the center position of the length. The deviation amounts up to are set to the following relationship, and the black region having the deviation amount equal to or more than the set value is excluded from the recognition target as the vehicle.

Δw / Hw ≦ 0.15 If the deviation amount of the position of the center of gravity in the horizontal direction, which is the parameter listed in (7), is less than or equal to a predetermined value, it is recognized as a vehicle that is the parameter listed in (8). The inclination of the principal axis of inertia of the target black area is detected, and when the inclination is extremely large, the black area is excluded from the recognition target as a vehicle.

FIG. 14 shows a binary image (A) when the preceding vehicle is in a short distance from the own vehicle, a binary image (B) when the preceding vehicle is in a long distance from the own vehicle, and a vehicle recognized as a vehicle. The binary image (C) in which the gymnastics is performed on the shadow of the side wall other than the target is shown.

As shown in FIGS. 14A and 14B, in the case of a black area recognized as a vehicle, the main axis of inertia (in the figure,
The line indicated by the solid line) is substantially horizontal.

On the other hand, as shown in FIG. 14C, in the case of a black region other than the object recognized as a vehicle, the principal axis of inertia determined from its shape has an inclination (θ) from the horizontal line. There is.

In this embodiment, the inclination (θ) of the principal axis of inertia is set to the following relationship, and the black area having the principal axis of inertia equal to or greater than the inclination is excluded from the recognition target as a vehicle.

Θ ≦ 10 ° In order to discriminate the black area to be recognized as a vehicle by using the above-mentioned parameters, in this embodiment, the processing based on the flowchart shown in FIG. 15 is executed.

In FIG. 15, the image picked up by the camera 1 is subjected to noise removal processing after binarization processing (S20).

In the binary image, the barycentric position of the black area is first detected, and it is determined whether or not the barycentric position is within a predetermined range (S21). In this case, the process using the parameters listed in (1) is executed. When it is determined that the position of the center of gravity is outside the predetermined range, the black region is excluded from the recognition target because it does not correspond to the recognition target as the vehicle,
Subsequent discrimination with respect to the black area is not performed.

If the position of the center of gravity is within the predetermined range in step S21, the aspect ratio of the black area in the binary image is detected, and it is determined whether the aspect ratio is less than or equal to the set value (S22). . In this case, the process using the parameters listed in (2) is executed. If it is determined that the aspect ratio is equal to or greater than the set value, the black region is excluded from the recognition target because it does not correspond to the recognition target as the vehicle, and the subsequent determination for the black region is not performed.

In step S22, when the aspect ratio is less than or equal to the set value, the actual area of the black area in the binary image and the area of the vehicle area corresponding to the rectangular area in contact with the black area are detected. It is determined whether the area ratio is less than or equal to the set value (S23). In this case, the process using the parameters listed in (3) is executed. If the area ratio is equal to or larger than the set value, the black region is excluded from the recognition target because it does not correspond to the recognition target as the vehicle, and the subsequent determination for the black region is not performed.

In step S23, if the area ratio is equal to or less than the set value, the uppermost coordinate position of the black area in the binary image is detected, and it is determined whether or not the position is equal to or more than the set value. (S24). In this case, the process using the parameters listed in (4) is executed. When it is determined that the coordinate position of the uppermost portion is equal to or greater than the set value, the black region is excluded from the recognition target because it does not correspond to the recognition target as the vehicle, and the subsequent determination is not performed.

In step S24, when the coordinate position is equal to or less than the set value, the actual area of the black area in the binary image is detected, and it is determined whether or not the actual area is within the set range ( S25). In this case, the processing using the parameters listed in (5) is executed. If it is determined that the actual area is not within the set range, the black area is excluded from the recognition target because it does not correspond to the recognition target as the vehicle, and the subsequent determination is not performed.

In step S25, if the actual area is within the set range, the horizontal length in the black area in the binary image is detected, and it is determined whether or not the horizontal length is within the set range. (S26). The processing in this case is
The processing using the parameters listed in (6) is executed.
If it is determined that the horizontal length is out of the set range, the black region is excluded from the recognition target because it does not correspond to the recognition target as the vehicle, and the subsequent determination is not performed.

In step S26, when the horizontal length is within the set range, the center position of the horizontal length and the center of gravity position in the black area in the binary image are detected, and the horizontal position between the positions is detected. It is determined whether the deviation in the direction is less than or equal to a predetermined value (S27). In this case, the process using the parameters listed in (7) is executed. When it is determined that the horizontal positional deviation is equal to or greater than the predetermined value, the black area is excluded from the recognition target because it does not correspond to the recognition target as the vehicle, and the subsequent determination is not performed.

In step S27, if the horizontal positional deviation is equal to or less than a predetermined value, the inclination of the inertial principal axis of the black area in the binary image is detected, and the inclination is equal to or less than the predetermined value. It is determined whether or not (S28). In this case, the processing using the parameters listed in (8) is executed. When the inclination of the principal axis of inertia is equal to or larger than the predetermined value, the black area is excluded from the recognition target because it does not correspond to the recognition target as the vehicle, and the subsequent determination is not performed.

As described in the above embodiments, the present invention is not limited to a vehicle as a recognition target, but it is also possible to target an obstacle, for example.

[0069]

As described above, according to the first aspect of the present invention, the barycentric position of the vehicle candidate area in the binary image is obtained, so that the erroneous detection component which is not the vehicle candidate area is simply processed. This makes it possible to exclude vehicles, which makes it possible to improve vehicle detection accuracy by not using complicated image processing.

According to the second aspect of the present invention, since the aspect ratio of the vehicle candidate area in the binary image is obtained, the erroneous detection component which is not the vehicle candidate area can be accurately excluded by a simple process. Therefore, the vehicle detection accuracy can be improved.

According to the third aspect of the present invention, the vehicle candidate region in the binary image is indexed by the area ratio that can be processed by simple image processing. The vehicle detection accuracy can be improved.

According to the fourth aspect of the present invention, it is possible to remove the erroneous detection component that does not correspond to the vehicle candidate area by using the coordinate height that can be easily calculated from the screen, so that the vehicle-likeness can be simplified. It can be checked, which can improve the detection accuracy of the vehicle.

According to the fifth aspect of the present invention, since the recognition target as a vehicle candidate can be determined only by using the actual area of the vehicle candidate area in the binary image, the vehicle detection accuracy can be improved by a simple process. You can

According to the sixth aspect of the present invention, the false detection component can be removed by checking the vehicle-likeness from the horizontal length of the vehicle candidate region in the binary image, so that only simple parameters are used. The processing can improve the vehicle detection accuracy.

According to the seventh aspect of the present invention, the vehicle-likeness can be checked by the horizontal length of the vehicle candidate area regardless of the size of the vehicle candidate area which changes depending on the distance to the photographed object. The vehicle detection accuracy can be improved without being affected by the inter-vehicle distance or the distance to the obstacle.

According to the invention described in claim 8, since only the deviation between the horizontal center position and the center of gravity position of the vehicle candidate region in the binary image is used, the erroneous detection component can be easily determined. Thus, the vehicle detection accuracy can be improved.

According to the ninth aspect of the present invention, since the inclination of the principal axis of inertia of the vehicle candidate area in the binary image is only checked, the erroneous detection component can be removed by a simple process, and the erroneous detection component can be removed. Since it is possible to determine a vehicle candidate area that does not include the, vehicle detection accuracy can be improved.

According to the tenth aspect of the present invention, since the erroneous detection component is detected using each parameter, the vehicle detection accuracy can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram for explaining a main part configuration used in a vehicle detection method according to the present invention.

FIG. 2 is a flowchart for explaining processing contents of a vehicle detection method according to the present invention.

FIG. 3 is a diagram showing a screen depicting a front image from the driver's seat side, which is captured by the camera shown in FIG. 1.

FIG. 4 is a diagram showing a screen depicting a state in which a black region in the image shown in FIG. 3 is expanded.

5 is a schematic diagram for explaining a method for expanding a black region on the screen of FIG.

FIG. 6 is a diagram showing a screen in a state where the screen shown in FIG. 3 is binarized.

FIG. 7 is a diagram showing a screen for performing noise removal processing on a binarized screen.

8 is a diagram showing a screen in a state where a black region of the screen shown in FIG. 7 is expanded.

9 is a diagram showing a screen in a state where a black area of the screen shown in FIG. 8 is reduced.

10 is a diagram showing a screen in a state where the black area of the screen shown in FIG. 9 is expanded to the original size.

11 is a diagram illustrating a screen in a state after noise removal processing is performed on the binary image shown in FIG.

12A and 12B are diagrams for explaining a method of obtaining an area ratio in the binary image shown in FIG. 6, where FIG. 12A shows a short distance and FIG. 12B shows a long distance.

13A and 13B are diagrams for explaining a method of obtaining a barycentric position in the binary image shown in FIG. 6, where FIG. 13A is a black area corresponding to a vehicle candidate area and FIG. 13B is not a vehicle candidate. Each black area is shown.

14A and 14B are diagrams for explaining a method for obtaining an inertial principal axis in the binary image shown in FIG. 6, where FIG. 14A is a black region corresponding to a vehicle candidate region at a short distance, and FIG. A black area corresponding to a vehicle candidate area at a long distance is shown, and (C) shows a black area not corresponding to a vehicle candidate.

FIG. 15 is a flowchart for explaining the procedure of the vehicle detection method according to the present invention.

FIG. 16 is a schematic diagram showing a vehicle and a shadow in the vicinity of the vehicle during traveling.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 camera 2 display device 3 image processing device 4 storage device 5 CPU 6 constituting the arithmetic unit storage device

Claims (10)

[Claims] 1. A vehicle detection method for binarizing a road image captured by a camera, extracting a vehicle candidate region from a black region of the binarized image, and recognizing a vehicle on the road based on the vehicle candidate region. In the vehicle detection method, the position of the center of gravity of the vehicle candidate region is detected, and when the position of the same center of gravity is outside a predetermined range, the vehicle candidate region is excluded from recognition targets of the vehicle. 2. A vehicle detection method for binarizing a road image captured by a camera, extracting a vehicle candidate region from a black region of the binarized image, and recognizing a vehicle on the road based on the vehicle candidate region. In the vehicle detection method, the aspect ratio of the vehicle candidate area is detected, and when the aspect ratio is equal to or greater than a set value, the vehicle candidate area is excluded from recognition targets of the vehicle. 3. A vehicle detection method for binarizing a road image captured by a camera, extracting a vehicle candidate region from a black region of the binarized image, and recognizing a vehicle on the road based on the vehicle candidate region. In, the actual area of the vehicle candidate area and the area of the rectangular area in contact with the vehicle candidate are detected, and when the area ratio of the actual area to the area of the rectangular area is less than or equal to a set value, the vehicle candidate area is set to the vehicle. The vehicle detection method is characterized by excluding it from the recognition target of the vehicle. 4. A vehicle detection method for binarizing a road image captured by a camera, extracting a vehicle candidate region from a black region of the binarized image, and recognizing a vehicle on the road based on the vehicle candidate region. In the vehicle detection, the top coordinate height of the vehicle candidate area is detected, and when the top coordinate height is equal to or more than a set value, the vehicle candidate area is excluded from recognition targets of the vehicle. Method. 5. A vehicle detection method for binarizing a road image captured by a camera, extracting a vehicle candidate region from a black region of the binarized image, and recognizing a vehicle on the road based on the vehicle candidate region. In the vehicle detection method, the actual area of the vehicle candidate area is detected, and when the actual area is out of a set range, the vehicle candidate area is excluded from recognition targets of the vehicle. 6. A vehicle detection method for binarizing a road image captured by a camera, extracting a vehicle candidate region from a black region of the binarized image, and recognizing a vehicle on the road based on the vehicle candidate region. In the vehicle detection method, the horizontal length of the vehicle candidate area is detected, and when the horizontal length is out of a set range, the vehicle candidate area is excluded from recognition targets of the vehicle. 7. The vehicle detection method according to claim 6, wherein the upper limit value and the lower limit value of the setting range are set to be smaller as the image height of the vehicle candidate region is higher. . 8. A vehicle detection method for binarizing a road image captured by a camera, extracting a vehicle candidate region from a black region of the binarized image, and recognizing a vehicle on the road based on the vehicle candidate region. In the horizontal length direction center position of the vehicle candidate region and the center of gravity position of the vehicle candidate region are detected, when the horizontal deviation between the center position and the center of gravity position is a predetermined value or more, the vehicle candidate region is selected. A vehicle detection method, characterized in that the vehicle is excluded from recognition targets. 9. A vehicle detection method for binarizing a road image captured by a camera, extracting a vehicle candidate region from a black region of the binarized image, and recognizing a vehicle on the road based on the vehicle candidate region. In the vehicle detection method, the inertial principal axis of the vehicle candidate area is detected, and when the inclination of the inertial principal axis is equal to or larger than a predetermined value, the vehicle candidate area is excluded from recognition targets of the vehicle. 10. A vehicle detection method for binarizing a road image captured by a camera, extracting a vehicle candidate region from a black region of the binarized image, and recognizing a vehicle on the road based on the vehicle candidate region. In the above, the position of the center of gravity of the vehicle candidate area is detected, the aspect ratio is detected in the vehicle candidate area, the area ratio between the actual area of the vehicle candidate area and the area of the rectangular area in contact with the vehicle candidate area is detected, and the maximum of the vehicle candidate area is detected. Detection of the upper coordinate height, detection of the actual area of the vehicle candidate area, detection of the horizontal length of the vehicle candidate area, detection of the horizontal deviation between the horizontal center of the vehicle candidate area and the center of gravity of the candidate area. , The main axis of inertia of the vehicle candidate area is continuously detected, and when each detection result matches the vehicle candidate exclusion condition set for these detection results, the vehicle candidate area is recognized by the vehicle. Exclude from A vehicle detection method characterized by: