CN102610104B - Onboard front vehicle detection method - Google Patents

Abstract

The invention discloses a vehicle-mounted vehicle detection method in front, comprising the following steps: 1) obtaining continuous images through a camera, and transmitting them to a computer for processing; 2) setting a specific detection area R1 on the image to be detected, and using The vehicle's bottom shadow feature and license plate feature detect the vehicle in front; 3) use the license plate feature of the vehicle to determine the license plate of the vehicle in front. The invention can be used on a moving camera or a still camera, and has small calculation amount, good real-time performance, certain adaptability to different light intensities, high recognition rate, and can meet general application requirements.

Description

A vehicle-mounted vehicle detection method in front

technical field

The invention relates to a vehicle-mounted vehicle detection method in front, and belongs to the technical field of application of image processing in intelligent transportation.

Background technique

At present, the method of detecting the vehicle in front is generally processed when the camera is fixed, and the processing needs to process the entire image pixel, plus preprocessing, feature extraction, detection, and identification methods are more complicated, and the amount of calculation is very large. Large, the burden on the processor is relatively large, and the general processor cannot guarantee the speed of image processing while completing other tasks. Otherwise, a relatively high-end processor is required, which increases the cost of image processing.

Contents of the invention

The technical problem to be solved by the present invention is to provide a front vehicle detection method, which can be used on a motion camera, has a small amount of calculation, and can meet real-time requirements in actual use.

In order to solve the problems of the technologies described above, the technical solution adopted in the present invention is:

A vehicle-mounted vehicle detection method ahead, characterized in that, comprising the following steps:

1) Obtain continuous images through the camera and send them to the computer for processing;

2) Set a specific detection area R1 on the image to be detected, and use the bottom shadow feature and license plate feature of the vehicle to detect the vehicle in front;

3) Using the license plate features of the vehicle to determine the license plate of the vehicle in front.

The aforementioned vehicle detection method in front is characterized in that: in the step 2), the specific detection region R1 has an image width of W/4 to W*3/4 and a height of H/2 to H of the image, Where W and H are the width and height of the image.

The aforementioned vehicle detection method in front is characterized in that: in the process of on-site video detection, through the numerical statistics and analysis of shadow pixels in the image, it is judged whether there is a vehicle in front, and when it is judged that there are consecutive cars in front that reach the threshold frame number, We think that there is car ahead surely.

The aforementioned vehicle detection method in front is characterized in that: the numerical statistics and analysis method of image shadow pixels are as follows: if the gray values of the three channels of pixels R, G, and B are lower than the threshold, and the gray values of the three channels are consistent with the area gray values If the difference of 1/3 of the mean value is greater than the threshold, the pixel is marked as white as the foreground; otherwise, it is marked as black as the background, and a single erosion and expansion process is performed on the processed image to obtain the shadow area closest to the bottom edge of the image The shape and statistical characteristics of R2.

The aforementioned vehicle detection method in front is characterized in that: in the shape and statistical features of the shadow region R2, the shape feature refers to the width and height of the rectangle circumscribing the shadow region R2 and its position in the image, and the statistical feature refers to the white pixels occupying the entire The percentage of the area of the circumscribed rectangle of the shaded area R2, when the shaded pixels marked as white meet the shape characteristics and statistical characteristics, that is, the width is within the threshold range, the absolute value of the difference between the horizontal position and the W/2 of the image is less than the threshold, and the white pixel If the percentage is greater than the threshold, the shadow area is considered temporarily as the shadow area of the vehicle, there is a car in the image, and the position of the vehicle is directly in front of the camera.

The aforementioned vehicle detection method in front is characterized in that: when the acquisition of marked white shadow pixels meets the conditions, and the number of consecutive frames of images satisfying the conditions is equal to the threshold, a rectangle located above the shadow area is taken as the feature area, and the width is The width and height of the shadow area are determined by the height of the camera and the resolution of the image. In the feature area, the gray average value of the original image is calculated as the identification feature of the vehicle, and the continuous lines greater than a certain threshold are projected through the horizontal projection of the vertical edge. Whether the number is greater than the set value, so as to judge whether there is a license plate.

The foregoing vehicle detection method in front is characterized in that: after the vehicle is detected, the identification feature of the feature area of the current vehicle is compared with the identification feature of the feature area calculated last time. The false detection rate and repeated capture rate increase the probability of identifying real non-repeated vehicles.

The beneficial effect of the invention is that it can be used on a moving camera or a still camera, and has small calculation amount, good real-time performance, certain adaptability to different light intensities, high recognition rate, and can meet general application requirements.

Description of drawings

Fig. 1 is a flow chart of the video detection vehicle in front provided by the present invention.

Detailed ways

Below in conjunction with accompanying drawing and embodiment the present invention is further described:

In order to solve the low recognition rate of the moving camera platform in video detection and the real-time performance of recognition, the present invention provides a simple method for detecting vehicles ahead. Describe in detail below in conjunction with accompanying drawing:

As shown in Figure 1, the flow chart of the video detection front vehicle provided by the present invention:

1. The hardware used in the present invention is a camera and a DSP computer, wherein the resolution of the camera is 1360(W)*1024(H). First, binarize part of the image area, select the W/4 to W*3/4 part of the image and the H/2 to H part as the image processing area R1, and in the area R1, the gray value of each pixel of the image is determined by three channels. If the difference between the intensity value and 1/3 of the regional gray value is greater than the threshold (15) and the gray values of the three channels are all less than the threshold (80) to select the shadow pixel in the image, if it is satisfied, it will be marked as white as the foreground, and the other Pixels that do not meet the condition are marked black, serving as the background.

2. Corrosion and expansion are performed on the areas described in 1 once to eliminate some of the noise points and voids that appear after binarization in 1.

3. Obtain from bottom to top the shadow area of the first shadow pixel circumscribed rectangle R2 that satisfies the threshold condition as the actual shadow area, and the proportion of white pixels in the area to the entire area is greater than the set threshold of 0.5, and the shadow If the distance between the center point of the area and the vertical centerline of the image is less than the threshold (40), it returns true, otherwise it returns false.

4. If the area in the image meets the above conditions, it is determined that there is a car ahead, and the continuous number of the vehicle is increased by one, otherwise the continuous number is cleared to prevent false detection.

5. After several frames of continuous monitoring, judge whether the number of consecutive frames with cars is equal to the threshold (8), if so, go to step 6, otherwise go to step 8.

6. Take the rectangular area R3 above the circumscribed rectangle R2 obtained in step 3, which is equal in width to R2 and has a height of 0.2*H/H1, as the characteristic area of the vehicle, where H is the vertical resolution of the image, and H1 is the installation height of the camera , the unit of H1 is meter. Take the weighted gray mean value of the three channels of all pixels in R3 as the feature value of the frame image. According to experience, we know that there are generally license plates in this area, and the texture of the license plate is richer than other parts of the vehicle. Perform SOBEL vertical edge detection processing on R3 and horizontal projection of the image after vertical edge detection, and determine the continuous lines whose projection is greater than the threshold (60) Whether the number is greater than the set threshold (20), and whether the difference between its eigenvalue and the last obtained eigenvalue is greater than the threshold (80). If it is satisfied, it is determined that there is a car and it is not the same vehicle as the last snapshot, go to step 7, otherwise go to step 8.

7. Save the image of the current frame as the image of the car.

8. Read the next frame of image and execute from step 1.

The above has disclosed the present invention with preferred embodiments, but it is not intended to limit the present invention, and all technical solutions obtained by adopting equivalent replacement or equivalent transformation methods fall within the protection scope of the present invention.

Claims (3)

1. a vehicle-mounted front vehicles detection method, is characterized in that, comprises the following steps:

1) by video camera, obtain continuous image, and reach in computing machine and process;

2) particular detection region R1 is set on image to be detected, and utilizes bottom shadow feature and the vehicle license plate characteristic of vehicle to detect front vehicles;

3) and utilize the vehicle license plate characteristic of vehicle to determine the car plate of front vehicles,

In described step 2) in, the width that described particular detection region R1 is image is at W/4 to W*3/4, and height is at the H/2 of image to H, and what wherein W and H were image is wide and high; In the process that live video is detected, whether by the numerical statistic of image shadows pixels and analysis, differentiating the place ahead has vehicle, when differentiate the place ahead have continuously car reach threshold value once frame number time think that there is car in certain the place ahead; The numerical statistic of image shadows pixels and analytical approach are, if pixel R, G, the three-channel gray-scale value of B is lower than threshold value two, and 1/3 difference of three-channel gray-scale value and area grayscale average is greater than threshold value three, this pixel is labeled as to white as prospect; Otherwise be labeled as black as a setting, and the image after processing is done to single corrosion expansion process, to obtain shape and the statistical nature of the shadow region R2 nearest from image base.

2. front vehicles detection method according to claim 1, it is characterized in that: in the shape and statistical nature of shadow region R2, shape facility refers to the wide height of shadow region R2 boundary rectangle and the position in image thereof, statistical nature refers to that white pixel accounts for the number percent of whole shadow region R2 boundary rectangle area, when being labeled as white direct-shadow image vegetarian refreshments, meet shape facility and statistical nature, be that width is in threshold value seven scopes, the W/2's of horizontal level and image is poor that absolute value is less than threshold value four, and white pixel number percent is greater than threshold value five, temporarily think that this shadow region is vehicle shadow region, in image, there is car, and the dead ahead of the position of vehicle in video camera.

3. front vehicles detection method according to claim 2, it is characterized in that: when obtaining, be labeled as shade of white pixel and satisfy condition, and there is the continuous frame number of the image satisfying condition to equal threshold value one, get a rectangle that is positioned at top, shadow region as characteristic area, width is shadow region width, the height that height is placed by video camera and the resolution of image determine, in characteristic area, ask for the average gray of original image, identification characteristics as vehicle, and whether the continuous line number that the horizontal projection by vertical edge is greater than certain threshold value six is greater than setting value, with this, judge whether to exist car plate.