JP3324590B2 - Road object 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 road object detecting apparatus for detecting an object on a road such as a traveling vehicle, a motorcycle or a falling object by imaging a road with a video camera.

[0002]

2. Description of the Related Art Numerous proposals have been made as conventional devices of this kind, such as a road object detecting device disclosed in Japanese Patent No. 2913882. FIG. 4 is a diagram illustrating a generalized apparatus of this type in the related art. 1 in the figure
Is a video camera for imaging a road, 2 is an analog / digital converter (hereinafter abbreviated as A / D converter) for converting a video signal output from the video camera 1 into digital image data, and 3 is superimposed on the video signal. A synchronizing signal separating unit 4 for separating and outputting the synchronizing signal is a clock generating unit for outputting a clock synchronized with the synchronizing signal output from the synchronizing signal separating unit 3, and a 5 is output from the A / D converter 2. Median value of inputting image data and outputting, for example, the median luminance value of a mask size of 3 × 3 pixels, that is, the fifth luminance value out of the total of 9 pixels of the current scanning position pixel and 8 peripheral pixels as the current image data Filter. Reference numeral 9 denotes an edge emphasizing unit for inputting image data output from the median filter 5 and performing edge emphasizing processing.
A Laplacian filter that outputs as a current image data the product-sum result of nine pixels, where the multiplier of the central pixel of three pixels, that is, the multiplier of the current scanning position pixel is 8, and the multiplier of the peripheral eight pixels is 1, or a source using another multiplier. Means such as a bell filter are generally used. Reference numeral 10 denotes a binarizing unit which binarizes the image data output from the edge emphasizing unit 9 with a predetermined threshold value and outputs so-called binary image data in which one bit indicates white or black. This is a contraction unit that receives the binary image data output by the binarization unit 10 and performs contraction processing of the binary image. For example, the number of white pixels in 25 pixels of a mask size of 5 × 5 pixels is a predetermined contraction threshold. In the following case, black is output as the current binary image data. Reference numeral 12 denotes an expansion unit for inputting the binary image data output from the contraction unit 11 and performing expansion processing of the binary image.
If the number of white pixels in 25 pixels of × 5 pixels is equal to or larger than a predetermined expansion threshold value, white is output as the current binary image data. Reference numeral 13 denotes an object detection unit that inputs the binary image data output from the expansion unit 12 and performs an object detection process.

Next, the operation will be described. The image data output from the A / D converter 2 is multi-valued image data having a halftone luminance of about 8 bits per pixel, and includes noise components. The median filter 5 is a well-known means as a filter for removing noise without damaging the edges of the image. The edge enhancement unit 9 enhances a change point, that is, an edge of the luminance of the image, based on a weighted difference between the peripheral pixel and the current pixel, or between the peripheral pixels, or the like. Thus, an edge image with less noise can be obtained. Binarization unit 1
0 is, for example, 1 for the edge portion of the image, and 0 for the other portions. 1 corresponds to white and 0 corresponds to black. Binary image data is 1
Since the data amount is 1/8 of that of the multi-valued image data of 8 bits per pixel, the subsequent processing load is reduced.
The contraction unit 11 reduces noise of the binary image such as removal of isolated white pixels by performing contraction processing. The expansion unit 12 performs a so-called hole filling such as replacing an isolated black pixel in the white pixel group with a white pixel by performing an expansion process. In this way, the object detecting unit 13 obtains the binary image data which is easy to process the image on the road. Object detector 13
Then, the edge of the image, that is, the pixel whose data is 1 is searched from the binary image data, and the shape is recognized by performing tracking processing to detect the road object. For example, in the case of a vehicle, the vehicle is recognized as a vehicle from the feature that there is a horizontal rectangular number plate near the center of the lower part of the overall horizontal rectangular shape. In the case of a two-wheeled vehicle, it is recognized as a two-wheeled vehicle because it has a substantially round human head at the top in the overall shape of a substantially rectangular rectangle. In the case of shapes such as falling objects other than vehicles and motorcycles, there are various types such as rod-shaped, plate-shaped or cubic cardboard. Recognition becomes possible. Therefore,
The better the detection performance is, the more the edge line is not broken, and the worse the detection performance is, the more the broken line is. In the prior art, the median filter 5, the edge enhancement units 9, 2
The coefficients, threshold values, and mask sizes of the respective units of the value conversion unit 10, the contraction unit 11, and the expansion unit 12 are fixed to predetermined values.

[0004]

The conventional road object detecting apparatus has the above-described configuration, and the coefficients, threshold values, and mask sizes of the respective parts are commonly applied to one entire image screen. Therefore, the difference between the size of the visual field per pixel and the brightness gradient at the edge part due to the distance from the road object to the video camera, that is, objects such as vehicles traveling far away are small images, and objects such as vehicles traveling close. There is a problem in that, when an object is detected based on edge data, the image is easily affected by the object detection performance, because an image having a small difference in luminance between pixels adjacent to the edge portion is obtained.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and has as its object to provide a road object detecting apparatus having stable detection performance which is hardly affected by the distance between the road object and a video camera. I have.

[0006]

A road object detecting apparatus according to a first aspect of the present invention includes a video camera for capturing an image of a road, and a video camera.
The video signal output from the video camera is converted to digital image data.
Analog / digital converter for converting data
Outputs a clock synchronized with the synchronization signal superimposed on the signal.
Clock generator and the clock generator
Input the clock and synchronization signal to
A scanning position detector for sequentially outputting information on the scanning position of the
A predetermined coefficient corresponding to the image data based on the inspection position information
Parameter memory for sequentially outputting
Pixel pitch of image data based on coefficient output by memory
And a resolution changer that changes the
To emphasize the point of change in image brightness
Edge enhancement unit that performs edge enhancement processing, and this edge enhancement unit
Binarizing the image data output from the printer with a predetermined threshold value 2
The binarizing section and the binarized image data from the binarizing section.
To perform a contraction process and an expansion portion to perform an expansion process
With respect to the contracted portion, the scanning position is a distant position
In the case of, the mask pixel size is smaller than that in the vicinity position,
A cell that replaces the center pixel with black from white as it is farther away
Reduce the number of white pixels other than the center pixel in the size
A first parameter memory;
When the position is far, the mask pixel is smaller than when it is near.
Size, and set the center pixel from black to white as it is farther away.
Increase the number of white pixels other than the center pixel within the mask size to be replaced.
A second parameter memory for thresholding and a binarizing section
Detect objects existing on the road from the binarized image data
And an object detecting means for outputting.

[0007] The second road object detection apparatus according to invention, the road
Video camera and the output of this video camera
To convert video signals into digital image data.
Analog / digital converter, superimposed on video signal
Clock output that outputs a clock synchronized with the synchronization signal
Synchronous with the raw part and the clock output from this clock generator
Information on the current scanning position of the video camera by inputting a signal
Scanning position detector that sequentially outputs data, and supports scanning position information
Parameter memory for sequentially outputting predetermined coefficients
Multiply the image data by the coefficient output from the parameter memory
Perform edge enhancement processing to enhance the change point of the image brightness.
Edge enhancement section and the image data output from the edge enhancement section.
Binarization unit for binarizing data with a predetermined threshold value, and the binarization
For performing the contraction process on the binarized image data from the
A contraction section and an expansion section for performing an expansion process.
On the other hand, when the scanning position is a distant position, it is closer to a near position.
Smaller pixel size, the farther away the center
The center pixel in the mask size that replaces pixels from white to black
The first parameter memo to reduce the number of white pixels other than
And when the scanning position is far away from the expanding portion,
Set the mask pixel size smaller than that at the side position, and
Mask size to replace the center pixel from black to white as in
The second pattern to increase the number of white pixels other than the central pixel
Parameter memory and binarized image data output by the binarization unit.
Object detecting means for detecting an object existing on the road from the data.

[0008]

[0009]

[0010]

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiment 1 FIG. 1 is a view showing a first embodiment according to the present invention, in which 1 to 5 and 9 to 13 are the same as those shown in FIG. Reference numeral 6 denotes a scanning position detecting unit which inputs a clock output from the clock generating unit 4 and a synchronizing signal output from the synchronizing signal separating unit 3, and sequentially outputs current scanning position information of the video camera 1, and 7a denotes a coefficient for each scanning position. Is stored in advance, and a parameter memory for inputting the scanning position information and sequentially outputting the coefficients is provided. The parameter memory 8 changes the resolution of the image data output from the median filter 5 based on the coefficients output from the parameter memory 7a. This is a resolution changing unit. L1 is a line for supplying a clock from the clock generation unit 4 to the resolution changing unit 8, L2 is a line for supplying image data from the median filter 5 to the resolution changing unit 8,
L3 is a line for supplying a coefficient from the parameter memory 7a to the resolution change unit 8, and L4 is a line for supplying the image data after the resolution change from the resolution change unit 8 to the edge enhancement unit 9.

FIG. 2 is a diagram showing an example of the configuration of the resolution changing section 8, in which L1, L2, L3 and L4 are the same as those shown in FIG. Reference numeral 14 denotes a counter which counts down the coefficient value supplied from the parameter memory 7a in synchronization with the clock output from the clock generator 4 and outputs a 0 signal when the count reaches 0. Reference numeral 15a denotes a counter which is synchronized with the clock output from the clock generator 4. Median filter 5
A register 15b which loads all the image data supplied from the register 15 and sequentially outputs the image data. A register 15b receives the image data loaded in the register 15a and outputs 0 from the counter 14.
A register that loads the image data by a signal and sequentially outputs the image data.

The coefficient output from the parameter memory 7a is
For example, a small value such as 0 is output when the scanning position is a position of an image far from the video camera 1, and a large value such as 4 is output when the scanning position is a position of a near image. This coefficient is 4
In the case of, four identical pixel data are successively replaced and output, and in the case of 0, the image data output by the median filter 5 is output as it is. In other words, an image closer in appearance is equivalent to making the quantization pitch of the video signal coarser, which means that the resolution of the image has been changed.

An image picked up by the video camera 1 is larger or smaller than the object image at a distance from the video camera 1. On the other hand, since the image data is obtained by quantization by the A / D conversion unit 2 with a clock having a constant period, when attention is paid to the edge portion of the image, the edge portion of the closer object image is formed of more pixels. For this reason, in the output image of the median filter 5, the brightness difference between adjacent pixels is smaller at the edge portion of the object image located closer to the edge portion, and it becomes difficult for the edge enhancement portion 9 to enhance the edge. By changing the resolution of the image data in step (1), a nearby object image becomes image data in which the luminance difference between adjacent pixels is large, so that a sufficient edge-emphasized image can be obtained and stable object detection processing can be performed.

Embodiment 2 FIG. 3 shows a second embodiment of the present invention.
Is the same as that shown in FIG. Reference numeral 7b denotes a parameter memory for inputting the scanning position information output from the scanning position detection unit 6 and sequentially supplying a previously stored edge enhancement processing coefficient to the edge enhancement unit 9, and 7c denotes a scanning position output from the scanning position detection unit 6. A parameter memory for inputting information and sequentially supplying a pre-stored binarization threshold value to the binarization unit 10; a reference numeral 7d denotes a contraction process which receives scan position information output from the scan position detection unit 6 and stores the scan position information in advance; A parameter memory for sequentially supplying coefficients to the contraction unit 11;
Is a parameter memory for inputting the scanning position information output by the controller and sequentially supplying an expansion processing coefficient stored in advance to the expansion unit 12.

In the second embodiment, the processing coefficient of the edge enhancement output from the parameter memory 7b is, for example, a mask size of 3 × 3 pixels when the video camera 1 is at a distant scanning position and a mask size of 3 × 3 pixels when the scanning position is near. Is a 5 × 5 pixel. The multiplier of 5 × 5 pixels for edge enhancement at a nearby position is, for example, 16 for the central pixel, 0 for the neighboring 8 pixels,
Further, the outer adjacent 16 pixels are set to -1. The processing coefficient of this near position is 3 × 3 of the far position.
Since a weighted difference between the adjacent pixel and the next pixel is calculated as compared with the mask size of the pixel, the edge portion of the object image is easily emphasized even when the luminance difference between the adjacent pixels is small.

The binarization threshold value output from the parameter memory 7c has a higher value at a scanning position of a bright portion in the field of view of the video camera 1, and has a lower value at a scanning position of a dark portion. Since the image data is output, stable binary image data can be obtained even if there is variation in brightness in the screen.

The shrinkage processing coefficient output from the parameter memory 7d is such that the mask size is 3 × 3 pixels when the scanning position of the video camera 1 is far, and 5 × when the scanning position is near.
The number of pixels is five. Further, by making the contraction threshold value, that is, the number of white pixels in the mask for replacing the central pixel with black, smaller at a farther position, a contraction process in which edges of a distant small image are less likely to be erased, that is, from the video camera 1, It is possible to perform stable noise removal processing that is less affected by distance.

Further, the expansion processing coefficient output from the parameter memory 7e is such that when the scanning position of the video camera 1 is far, the mask size is 3 × 3 pixels, and when the scanning position is close, the mask size is 5 × 3 pixels.
The number of pixels is five. By increasing the expansion threshold value, that is, the number of white pixels in the mask that replaces the central pixel with white, as the distance to the position increases, the expansion processing in which the inside of the edge of the distant small image is less likely to be collapsed, that is, the video camera 1 A stable filling process that is hardly affected by the distance from the object can be performed.

[0020]

According to the first aspect of the present invention, the scanning position detecting section
The scanning position of the video camera is detected, and based on the position information
Parameter method for sequentially outputting the processing coefficients for changing the resolution
Change the resolution of the image by the moly and its processing factor
With a resolution changing unit, it is affected by the distance of road objects.
Stable detection performance and scanning position information
The processing coefficients for contraction and expansion based on
The first and second parameter menus sequentially supplied to each
With a molybdenum, the object on the road
Stable denoising and holes independent of distance
The filling process can be performed and the detection performance is stabilized .

According to the second invention, the scanning position information
The coefficient output from the parameter memory based on the
To be supplied to the edge emphasizing unit as a key processing coefficient
Configuration, so that it is not affected by the distance of road objects.
Edge-enhanced image, and based on scanning position information
The contraction and expansion coefficients
First and second parameter memories sequentially supplied to the
The distance of the object on the road to the binarized image data
Unaffected and stable noise removal and fill processing
And the detection performance is stabilized .

[0022]

[0023]

[0024]

[Brief description of the drawings]

FIG. 1 is a diagram showing Embodiment 1 of a road object detection device according to the present invention.

FIG. 2 is a diagram showing a configuration example of a resolution changing unit of the road object detection device according to the present invention.

FIG. 3 is a diagram showing Embodiment 2 of a road object detection device according to the present invention;

FIG. 4 is a diagram showing a conventional road object detection device.

[Explanation of symbols]

Reference Signs List 1 video camera, 2 A / D conversion section, 3 synchronization signal separation section, 4 clock generation section, 6 scanning position detection section, 7
a, 7b, 7c, 7d, 7e Parameter memory, 8
Resolution change unit, 9 Edge enhancement unit, 10 Binarization unit, 1
1 contraction section, 12 expansion section, 13 object detection section.

──────────────────────────────────────────────────続 き Continued on the front page (58) Fields surveyed (Int. Cl. ⁷ , DB name) G08G 1/00-1/054 G06T 1/00-5/20

Claims (2)

(57) [Claims] 1. A video camera for capturing an image of a road and a video signal output from the video camera
An analog / digital conversion unit for converting the image data into a video signal; and a clock synchronized with a synchronization signal superimposed on the video signal.
A clock generator for outputting a lock, a clock output by the clock generator, and the synchronization signal
To input information on the current scanning position of the video camera.
And a scanning position detecting unit for sequentially outputting the image data corresponding to the image data based on the scanning position information.
A parameter memory for sequentially outputting a predetermined coefficient, and the image memory based on the coefficient output from the parameter memory.
A resolution changing unit for changing the pixel pitch of the image data; and an image data output from the resolution changing unit.
Edge strength to perform edge enhancement processing to emphasize the change point of luminance
The tone portion and the image data output from the edge emphasizing portion have a predetermined threshold value.
A binarizing unit for binarizing the image with a value, and a binarizing process for the binarized image data from the binarizing unit.
And a dilation unit that performs dilation processing . When the scanning position is a distant position, the dilation unit is located near the dilation unit.
The mask pixel size is smaller than the
Within the mask size to replace the center pixel from white to black from time to time
The first parameter that reduces the number of white pixels other than the central pixel
With respect to the meter memory and the inflated portion, when the scanning position is a distant position, it is near.
The mask pixel size is smaller than the
Within the mask size to replace the central pixel from black to white from time to time
To increase the number of white pixels other than the central pixel
A meter memory and the binarized image data output by the binarizing unit are used to place on a road.
An on-road object detection device , comprising: an object detection unit that detects an existing object . 2. A video camera for capturing an image of a road, an analog / digital converter for converting a video signal output from the video camera into digital image data, and a clock synchronized with a synchronization signal superimposed on the video signal A clock generating unit that outputs a clock output by the clock generating unit and a synchronizing signal, and sequentially outputs information on a current scanning position of the video camera; A parameter memory for sequentially outputting predetermined coefficients to be processed, an edge emphasizing section for performing an edge emphasizing process for multiplying the image data by the coefficients output from the parameter memory to emphasize a change point of the luminance of the image, and an output from the edge emphasizing section. A binarization unit for binarizing the image data to be binarized with a predetermined threshold value, and binarized image data from the binarization unit. And shrinking
And a dilation unit that performs dilation processing . When the scanning position is a distant position, the dilation unit is located near the dilation unit.
The mask pixel size is smaller than the
Within the mask size to replace the center pixel from white to black from time to time
The first parameter that reduces the number of white pixels other than the central pixel
With respect to the meter memory and the inflated portion, when the scanning position is a distant position, it is near.
The mask pixel size is smaller than the
Within the mask size to replace the central pixel from black to white from time to time
To increase the number of white pixels other than the central pixel
And a meter memory, the road object detection apparatus characterized by comprising from binarized image data binarizing unit outputs the object detecting means for detecting an object existing on the road, the.