JP2012150587A - Vehicle detection device and vehicle detection method - Google Patents

Abstract

There is provided a vehicle detection device capable of improving the accuracy of recognizing a vehicle even in a region where a proximity image is detected and an edge strength is reduced.
A vehicle detection apparatus 1 according to the present invention includes a gray scale unit 3 that converts an image obtained by capturing the surroundings of a host vehicle to gray scale, and an edge processing unit 4 that performs edge processing on the gray scaled captured image using a filter. A vehicle image cutout unit 5 that cuts out a vehicle image indicating the characteristics of the vehicle from the edge-processed captured image, a vehicle image division unit 6 that divides the cut out vehicle image into a plurality of divided regions, and The edge addition unit 7 for adding the edge strength of the upper pixel row to the edge strength of each pixel row constituting the divided region, and comparing the edge strength of the pixels constituting the divided region with a preset threshold value A vehicle recognition unit 8 that recognizes the presence or absence is provided.
[Selection] Figure 1

Description

  The present invention relates to a vehicle detection apparatus and a vehicle detection method for recognizing and detecting an image of a vehicle from a captured image obtained by imaging the periphery of the host vehicle.

  In recent years, the development of technology for detecting vehicles existing around the subject vehicle by analyzing images taken around the vehicle has been promoted, and technology for safe driving of vehicles such as automatic steering has been developed. Applied.

  Normally, as a method for recognizing a vehicle from a captured image, the captured image is edge-processed with a filter, and a portion of an edge strength that is equal to or higher than a preset threshold value is detected to identify the position of the preceding vehicle.

  As an example of a technique for detecting such a vehicle, there is a vehicle detection device disclosed in Patent Document 1, for example. In this vehicle detection device, the average luminance of the pixel columns extending in the vertical direction of the area where the vehicle is imaged is calculated, and then the difference between the i-th column and the i + 1-th column is calculated to obtain the difference in average luminance. . Then, all the differences in the average luminance of each column are calculated for the area where the vehicle is imaged, and the maximum average luminance difference exceeding the threshold value is obtained, and the position of the column that is the maximum value is determined as the position of the vehicle. It is detected as a horizontal position.

JP 2009-175846 A

However, in the edge processing using the current filter, the near image has a lower edge strength than the far image, and a sufficient edge strength cannot be obtained in the close portion of the captured image.

Therefore, the edge strength may not reach the threshold in the region where the proximity image is detected in the captured image, and there is a problem in that the vehicle recognition rate decreases. In addition, if the threshold value is lowered to solve this problem, the edge strength generally increases in the region where the far-field image is detected, so that even those that do not need to be detected are detected, and false detection increases. Another problem has arisen.

Furthermore, in the vehicle detection device disclosed in Patent Document 1, the area where the vehicle is imaged is divided into three areas, an upper divided area, an intermediate divided area, and a lower divided area. The problem described above could not be solved because it was not changed.

Therefore, the present invention has been proposed in view of the above-described circumstances, and can improve the accuracy of recognizing a vehicle even in a region where edge strength is reduced in a region where a proximity image is detected. An object of the present invention is to provide a vehicle detection device and a vehicle detection method.

  In order to achieve the above-described object, the invention according to claim 1 of the present application includes a gray scale unit that gray-scales a captured image captured around the host vehicle, and a filter for the gray-scaled captured image. Edge processing means for performing edge processing, vehicle image cutting means for cutting out a vehicle image showing the characteristics of the vehicle from the edge-processed captured image, and dividing the cut-out vehicle image vertically into a plurality of divided regions Vehicle image dividing means; edge adding means for adding edge strength of the upper pixel row to edge strength of each pixel row constituting the divided region in at least one of the plurality of divided regions; and Vehicle recognition means for recognizing the presence or absence of the vehicle by comparing the edge intensity of the constituent pixels with a preset threshold value is provided.

  The vehicle image dividing means of the invention according to claim 2 of the present application divides the vehicle image into an upper divided area, a middle divided area, and a lower divided area, and the edge adding means has at least each pixel row in the lower divided area. The edge strength of the upper pixel row is added to the edge strength.

  The edge addition means of the invention according to claim 3 of the present application does not add edge strength in the upper divided region, and adds edge strength of one pixel row to the edge strength of each pixel row in the middle divided region. In the lower divided area, the edge strengths of the pixel row one level up and the pixel row two levels above are added to the edge strength of each pixel row.

  The invention according to claim 4 of the present application is a gray scale step for converting a captured image obtained by capturing the periphery of the host vehicle to a gray scale, an edge processing step for performing an edge process on the gray scaled captured image using a filter, A vehicle image cutting step for cutting out a vehicle image showing the characteristics of the vehicle from the edge-processed captured image, a vehicle image dividing step for dividing the cut-out vehicle image into a plurality of divided regions, and In at least one of the divided regions, an edge addition step of adding the edge strength of the upper pixel row to the edge strength of each pixel row constituting the divided region, and the edge strength of the pixels constituting the divided region are preset. A vehicle recognition step of comparing the detected threshold value and recognizing the presence or absence of the vehicle.

  In the inventions according to claims 1 and 4 of the present application, since the edge strength of the upper pixel row is added to the edge strength of each pixel row constituting the divided region in at least one divided region, the edge strength is reduced. Even in the divided area, the accuracy of recognizing the vehicle can be improved.

  In the invention according to claim 2 of the present application, the vehicle image is divided into three parts, and at least in the lower divided region, the edge strength of the upper pixel row is added to the edge strength of each pixel row. The accuracy of recognizing a vehicle can be improved even in a lower divided region that becomes lower.

  In the invention according to claim 3 of the present application, the edge strength is not added in the upper divided region, and more edge strength is added from the middle divided region to the lower divided region. The accuracy of recognizing the vehicle can be improved, and erroneous detection can be prevented in a region where a distant image is detected.

It is a block diagram which shows the structure of the vehicle detection apparatus which concerns on one Embodiment of this invention. It is a flowchart which shows the procedure of the vehicle detection process by the vehicle detection apparatus which concerns on one Embodiment of this invention. It is a figure which shows an example of the image grey-scaled by the vehicle detection apparatus which concerns on one Embodiment of this invention. It is a figure which shows an example of the image edge-processed by the vehicle detection apparatus which concerns on one Embodiment of this invention. It is a figure which shows an example of the vehicle image cut out by the vehicle detection apparatus which concerns on one Embodiment of this invention. It is a figure which shows an example of the image divided | segmented into the several division area by the vehicle detection apparatus which concerns on one Embodiment of this invention. It is a figure for demonstrating the addition process of the edge strength by the vehicle detection apparatus which concerns on one Embodiment of this invention. It is a figure for demonstrating the addition process of the edge strength by the vehicle detection apparatus which concerns on one Embodiment of this invention. It is a figure for demonstrating the effect by the vehicle detection apparatus which concerns on one Embodiment of this invention.

  Hereinafter, an embodiment to which the present invention is applied will be described with reference to the drawings.

[Configuration of vehicle detection device]
FIG. 1 is a block diagram showing a configuration of a vehicle detection device according to the present embodiment.

  As shown in FIG. 1, a vehicle detection device 1 according to the present embodiment includes an imaging unit 2 that captures an image around the host vehicle, and a grayscale unit 3 that converts an image captured by the imaging unit 2 to grayscale. An edge processing unit 4 that performs edge processing of the grayscale captured image with a filter, a vehicle image cutting unit 5 that cuts out a vehicle image indicating the characteristics of the vehicle from the edge-processed captured image, and A vehicle image dividing unit 6 that divides the vehicle image vertically into a plurality of divided regions, and an edge adding unit 7 that adds the edge strength of the upper pixel row to the edge strength of each pixel row that constitutes the divided region; A vehicle recognizing unit 8 that recognizes the presence or absence of a vehicle by comparing the edge intensity of the pixels constituting the divided region with a preset threshold value, and a storage unit that stores a captured image captured at a predetermined sampling period It is equipped with a door.

  Here, the vehicle detection apparatus 1 according to the present embodiment is mounted on a vehicle as a part of various systems such as an automatic steering system and a drive recorder, and can be adapted to control of various systems by changing software. it can. In the present embodiment, as shown in FIG. 1, the case where the vehicle detection device 1 is configured by hardware is shown as an example. However, the configuration is not limited to hardware, for example, a CPU and its You may make it implement | achieve the function of the vehicle detection apparatus 1 with the software of the computer program for operating CPU. When performing real-time processing, the CPU in the video recording device executes the algorithm of the vehicle detection device 1 according to the present embodiment. When performing batch processing, the video stored in the recording unit 9 is read out. To analyze.

  The imaging unit 2 may be configured by one or a plurality of cameras and can capture an image that can detect the preceding vehicle and surrounding vehicles. Moreover, the camera installed in the drive recorder may be sufficient.

  The gray scale conversion unit 3 captures an image captured by the image capture unit 2 at a predetermined sampling period and stores it in the storage unit 9, and converts the stored color image to gray scale. At this time, any method such as a simple average method or an intermediate value method may be used as the gray scale method.

  The edge processing unit 4 performs edge processing with various filters on the grayscale captured image. As a filter used here, for example, a Sobel filter can be used, but other filters may be used as long as they can detect edges.

  The vehicle image cutout unit 5 cuts out an image of a portion showing the characteristics of the vehicle as a vehicle image from the edge-processed captured image. There are various methods for cutting out, and any method may be adopted.

  The vehicle image dividing unit 6 divides the cut-out vehicle image into a plurality of divided areas in a vertical direction at a predetermined ratio. In this embodiment, the vehicle image dividing unit 6 divides the vehicle image into three areas: an upper divided area, a middle divided area, and a lower divided area. This is described as an example. However, it is not always necessary to divide into three, and it may be divided into two or may be divided into four or more. In addition, although the division ratio can be arbitrarily changed, in the present embodiment, an example in which the upper division area is 30%, the middle division area is 40%, and the lower division area is 30% will be described as an example.

  The edge adding unit 7 adds the edge strength of the upper pixel row to the edge strength of each pixel row constituting the divided region in at least one divided region. For example, when the vehicle image is divided into three, the edge strength is not added in the upper divided region, and in the middle divided region, the edge strength of the upper pixel row is added to the edge strength of each pixel row, and the lower divided region In this case, the edge strengths of the upper and lower pixel rows are added to the edge strength of each pixel row.

  The vehicle recognition unit 8 compares the edge strength of the pixels constituting each divided region with a preset threshold value, and recognizes that the vehicle is present when the edge strength exceeds the threshold value, If the edge strength is less than or equal to the threshold, it is recognized that no vehicle is present.

[Vehicle detection processing procedure]
Next, the procedure of the vehicle detection process by the vehicle detection apparatus 1 according to the present embodiment will be described with reference to the flowchart of FIG.

  As shown in FIG. 2, in step S <b> 101, an image around the host vehicle is captured by the imaging unit 2. The captured image is captured at a predetermined sampling period and stored in the storage unit 9, and the color image stored in step S102 is converted into a gray scale as shown in FIG. At this time, any method such as a simple average method or an intermediate value method may be used as the gray scale method.

  Next, in step S103, edge processing is performed with various filters on the captured image that has been grayscaled. For example, when the Sobel filter is used as an example, the captured image subjected to the edge processing is an image as shown in FIG. In step S104, a portion showing the characteristics of the vehicle is cut out from the edge-processed captured image to obtain a vehicle image as shown in FIG.

  Next, in step S105, the cut-out vehicle image is divided into a plurality of divided regions vertically at a predetermined ratio. In the present embodiment, as shown in FIG. 6, the upper divided area A, the middle divided area B, and the lower divided area C are divided into three parts. It may be divided or more. Further, the division ratio can be arbitrarily changed. In the present embodiment, the upper division area A is 30%, the middle division area B is 40%, and the lower division area C is 30%.

  When the vehicle image is thus divided, in step S106, processing for adding the edge strength of the upper pixel row to the edge strength of each pixel row constituting the divided region is executed for at least one divided region. . For example, when the vehicle image is divided into three, the edge strength is not added in the upper divided region A, and in the middle divided region B, the edge strength of the pixel row one level higher is added to the edge strength of each pixel row. In the divided region C, processing is performed in which the edge strengths of the pixel row one level up and the pixel row two levels above are added to the edge strength of each pixel row. In this way, in the upper divided area A, the edge strength is not added, and more edge strengths are added as going from the middle divided area B to the lower divided area C. This is because the upper divided area A is a relatively distant image, so that the edge strength is detected strongly, while the lower divided area C is a close image, so the edge intensity is detected low.

  Therefore, the accuracy of recognizing the vehicle can be improved even in a region where the edge strength is low in the close-up image by adding more edge strengths as going to the lower divided region C.

Here, with reference to FIG. 7, edge strength addition processing in the middle divided region B will be described. As shown in FIG. 7, when attention is paid to the target row 1 among the pixel rows constituting the middle divided region B, the addition result of the target row 1 is an addition that is one higher than the target row 1 to the edge strength of the target row 1. The edge strength of row a is added. That is, the addition result of the target row 1 = the edge strength of the target row 1 + the edge strength of the addition row a (1)
It becomes. Further, the addition result of the target row 2 is obtained by adding the edge strength of the target row 1 that is one higher than the target row 2 to the edge strength of the target row 2. That is, the addition result of the target row 2 = the edge strength of the target row 2 + the edge strength of the target row 1 (2)
It becomes. The edge strength in the middle divided region B is calculated by repeatedly performing such edge strength addition processing in the middle divided region B.

Next, edge addition processing in the lower divided area C will be described with reference to FIG. As shown in FIG. 8, when attention is paid to the target row 1 among the pixel rows constituting the lower divided area C, the addition result of the target row 1 is an addition that is one higher than the target row 1 to the edge strength of the target row 1. This is the sum of the edge strength of row a and the edge strength of addition row b that is two above target row 1. That is, the addition result of the target row 1 = the edge strength of the target row 1 + the edge strength of the addition row a
+ Edge strength of addition row b (3)
It becomes. Also, the addition result of the target row 2 is obtained by adding the edge strength of the target row 1 that is one above the target row 2 and the edge strength of the addition row a that is two above the target row 2 to the edge strength of the target row 2 Will be. That is, the addition result of the target row 2 = the edge strength of the target row 2 + the edge strength of the target row 1
+ Edge strength of addition row a (4)
It becomes. The edge strength in the lower divided region C is calculated by repeatedly performing such edge strength addition processing in the lower divided region C.

  When the added edge intensity is calculated in each divided area in this way, in step S107, the edge intensity of the pixels constituting each divided area is compared with a preset threshold value. For example, since the edge strength is not added in the upper divided region A, the edge strength not added is compared with the threshold value, and the added edge strength and the threshold value are compared in the middle divided region B and the lower divided region C. Compare. Here, the threshold value is a common value in each divided region. When the edge strength exceeds the threshold, the vehicle is recognized as existing, and when the edge strength is equal to or lower than the threshold, the vehicle is recognized as not present. When the presence or absence of the vehicle is recognized in this way, the vehicle detection process by the vehicle detection device 1 according to the present embodiment ends.

[Effect of the embodiment]
As described above in detail, according to the vehicle detection device of the present embodiment, in at least one divided region, the edge strength of the upper pixel row is added to the edge strength of each pixel row constituting the divided region. Therefore, it is possible to improve the accuracy of recognizing the vehicle even in a divided region where the edge strength is low in the proximity image.

  Here, with reference to FIG. 9, the effect by adding edge strength is demonstrated. FIG. 9 shows the edge strength of one pixel row, FIG. 9 (a) shows the value before adding the edge strength, and FIG. 9 (b) shows the value after adding the edge strength. Show.

  As shown in FIG. 9, it can be seen that the edge strength increases as a whole after adding the edge strength. Thereby, the accuracy of recognizing the vehicle can be improved even in a divided region where the edge strength is lowered due to the proximity image or the like.

  Further, according to the vehicle detection device according to the present embodiment, the vehicle image is divided into three parts, and at least in the lower divided region, the edge strength of the upper pixel row is added to the edge strength of each pixel row. Thus, the accuracy of recognizing the vehicle can be improved even in the lower divided area where the edge strength becomes low.

  Furthermore, according to the vehicle detection device according to the present embodiment, the edge strength is not added in the upper divided region, and more edge strengths are added from the middle divided region to the lower divided region, so that the proximity image is detected. The accuracy of recognizing the vehicle can be improved in the region where the image is detected, and erroneous detection can be prevented in the region where the far-field image is detected.

  The above-described embodiment is an example of the present invention. For this reason, the present invention is not limited to the above-described embodiment, and even if it is a form other than this embodiment, as long as it does not depart from the technical idea according to the present invention, the design and the like Of course, various modifications are possible.

  The present invention can be used for techniques for safe driving of vehicles such as automatic steering.

DESCRIPTION OF SYMBOLS 1 Vehicle detection apparatus 2 Imaging part 3 Gray scale conversion part 4 Edge process part 5 Vehicle image cutout part 6 Vehicle image division part 7 Edge addition part 8 Vehicle recognition part 9 Storage part

Claims (4)

Gray scale conversion means for converting a captured image obtained by imaging the periphery of the host vehicle to gray scale;
Edge processing means for edge-processing the grayscale captured image with a filter;
Vehicle image cutting means for cutting out a vehicle image showing the characteristics of the vehicle from the edge-processed captured image;
Vehicle image dividing means for dividing the clipped vehicle image vertically into a plurality of divided regions;
Edge addition means for adding the edge strength of the upper pixel row to the edge strength of each pixel row constituting the divided region in at least one of the plurality of divided regions;
A vehicle detection device comprising vehicle recognition means for recognizing the presence or absence of a vehicle by comparing an edge intensity of pixels constituting the divided area with a preset threshold value.   The vehicle image dividing means divides the vehicle image into an upper divided area, an intermediate divided area, and a lower divided area, and the edge adding means at least adds the edge intensity of each pixel line to the edge intensity of each pixel line in the lower divided area. The vehicle detection apparatus according to claim 1, wherein edge strength is added.   The edge addition means does not add edge strength in the upper divided region, adds the edge strength of one pixel row to the edge strength of each pixel row in the middle divided region, and adds each edge strength in the lower divided region. The vehicle detection device according to claim 2, wherein the edge strength of one pixel row and the second pixel row is added to the edge strength of the pixel row. A gray scale step for converting the captured image obtained by capturing the surroundings of the host vehicle to gray scale;
An edge processing step of performing edge processing on the grayscale captured image with a filter;
A vehicle image cutting step of cutting out a vehicle image showing the characteristics of the vehicle from the edge-processed captured image;
A vehicle image dividing step of dividing the cut-out vehicle image vertically into a plurality of divided regions;
An edge addition step of adding the edge strength of the upper pixel row to the edge strength of each pixel row constituting the divided region in at least one of the plurality of divided regions;
The vehicle detection method characterized by including the vehicle recognition step which recognizes the presence or absence of a vehicle by comparing the edge strength of the pixel which comprises the said division area, and the preset threshold value.