CN103473567A - Vehicle detection method based on partial models - Google Patents

Abstract

The invention relates to the technical field of vehicle detection, in particular to a vehicle detection method based on a partial model. The method includes the following steps: extracting two parts from the vehicle object according to the degree of easy occlusion to form a vehicle model; using a mixed image template fused with multiple features to model the two parts respectively; using the training image to learn the two parts The corresponding mixed image templates and the image likelihood probabilities of these templates, while learning the probability distribution of the position and scale relationship between the two parts; detecting the candidates of the two parts from the test traffic image, using the position and scale relationship between the parts The scale relation combines vehicle candidates to achieve vehicle detection. The invention has the advantages of adapting to moderate vehicle deformation, various weather conditions, etc., and can be applied to deal with vehicle occlusion in complex traffic scenes.

Description

A Vehicle Detection Method Based on Partial Model

technical field

The invention relates to the technical field of vehicle detection, in particular to a vehicle detection method based on a partial model.

Background technique

Video-based vehicle detection technology is an important part of intelligent transportation systems, providing vehicle information for many applications; such as driver assistance systems, traffic video surveillance systems, etc. At present, the commonly used vehicle detection method is to use the motion information of the vehicle to detect the vehicle, such as the paper "Adaptive Vehicle Detector Approach for Complex Environments" published by Wu Bingfei and others in "IEEE Transactions on Intelligent Transportation Systems (IEEE Intelligent Transportation Transactions)" in 2012 (Adaptive vehicle detector in complex scenes) is to use motion information to detect vehicles. However, this method is not suitable for dealing with occlusions between vehicles, and it is difficult to accurately separate adjacent vehicles that occlude each other. And it is also not applicable due to the lack of motion information of vehicles in slow traffic. In addition, vehicle detection methods based on image features often use information such as vehicle edges and textures to detect vehicles, such as the paper "Learning Active BasisModel for Object Detection" published by Wu Yingnian et al. and Recognition” (learning activity base model for target detection and recognition), which uses edge information to detect vehicles. However, many methods based on image features do not consider vehicle occlusion, and are not suitable for dealing with occluded traffic scenes.

Contents of the invention

The technical problem to be solved by the present invention is to provide a vehicle detection method based on a partial model to realize vehicle detection in complex traffic scenes with occlusions.

The technical scheme that the present invention solves the problems of the technologies described above is:

It mainly includes part selection, part modeling, model learning and vehicle detection;

In the part selection, two parts (a) and (b) are extracted from the vehicle object according to the degree of easy occlusion to form the vehicle model; part (a) is an area that is not easy to be occluded near the window, Part (b) is the easily occluded area around the license plate;

In the part modeling, the two parts are modeled using a mixed image template fused with multiple features;

The model learning uses the training image to learn the mixed image template of the two parts and the position and scale relationship between the two parts;

The vehicle detection uses an iterative process to detect one or more vehicles in the test traffic image. In each iterative step, template matching is first used to detect the candidates of the two parts, and then the position and scale relationship between the two parts are used to Combine some candidates to achieve vehicle detection.

The described model learning includes the following steps:

Step S3-1, intercepting vehicle images from actual traffic images as training images, the number of training images is not less than 1;

Step S3-2, using the message mapping method to learn all the image blocks in the mixed image template corresponding to part (a) and part (b) and the mixed image template corresponding to part (a) and part (b) from the training image The image likelihood probability of ;

Step S3-3, using the training image to learn the probability distribution of the position and scale relationship between part (a) and part (b).

The described vehicle detection utilizes an iterative process to detect one or more vehicles from the test traffic image, including the following steps:

Step S4-1, using the mixed image template of the part (a) and the part (b) to carry out template matching to the test traffic image, extracting the candidates of the part (a) and the part (b);

Step S4-2, combining the candidates of part (a) and part (b) into one or more vehicle candidates by using the position and scale relationship of said part (a) and part (b), and calculating these vehicle candidates or the corresponding vehicle detection probability;

Step S4-3, extracting the vehicle candidate with the maximum vehicle detection probability;

Step S4-4, comparing the maximum vehicle detection probability in the step S4-3 with the vehicle detection threshold, if the vehicle detection probability is greater than or equal to the vehicle detection threshold, the corresponding vehicle candidate is a detected vehicle object; then Use the test traffic image of the detected vehicle object to perform the next iterative step, repeat steps S4-1, S4-2, S4-3, S4-4; if the vehicle detection probability is less than the vehicle detection threshold, the entire iterative process Stop, the vehicle detection process ends.

The mixed image template is composed of one or more image blocks, and the image blocks are divided into types such as edge blocks, texture blocks, color blocks and flatness blocks, and the image blocks included in the mixed image template can be one of these types or more;

The edge block is represented by the Gabor wavelet primitive in a specific direction; the texture block is represented by the gradient histogram of the image area; the color block is represented by the color histogram of the image area; the flatness block is represented by the superposition Value representation obtained from the response of the Gabor filter in one or more directions within the image region.

The image likelihood probability of the mixed image template of the part (a) or part (b) is:

$\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; \&Lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; q</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; \&Pi;</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}\frac{\mathrm{<span\; class="notranslate">\; exp</span>}<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; \&Lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span>}{{\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}}<span\; class="notranslate">\; ,</span>$

之间的距离，Z_ij是归一化常数。where p ₁ and p ₂ are the mixed image templates of the part (a) and part (b) respectively, N _i is the number _of image blocks in _pi , q(I) is a reference distribution, λ _ij is The coefficient of the jth image block, is the jth image patch and image in p _i

The distance between Z _ij is the normalization constant.

The vehicle detection probability is:

$\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \}</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; \&Lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; \&Lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; \&Pi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; \&Lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>$

和

分别为在车辆检测过程中被检测出的所述部分(a)和部分(b)的候选者所在的图像区域，

表示了一对部分(a)和部分(b)的候选者之间的位置和尺度关系的概率。where I is a test traffic image,

and

are respectively the image areas where the candidates of the part (a) and part (b) are detected during the vehicle detection process,

Denotes the probability of the location and scale relationship between a pair of candidates for part (a) and part (b).

The calculation steps of the vehicle detection threshold include:

First, utilize step S4-1, S4-2, S4-3 to detect vehicle from all described training images, and calculate corresponding vehicle detection probability;

Then, the vehicle detection threshold is estimated using the vehicle detection probabilities of all the training images.

The vehicle detection method of the present invention has the following advantages:

(1) In part selection, two parts (a) and part (b) are selected from the vehicle object according to the degree of easy occlusion to form the vehicle model. Part (a) is the area around the window that is not easy to be occluded, and part (b) is the area around the license plate that is easily occluded. This partial selection reduces the influence of the occlusion situation on part (a).

(2) In the partial modeling, the mixed image template is used to model part (a) and part (b), and the mixed image template combines a variety of vehicle features, including edge, texture, color and flatness, etc., which improves the The detection accuracy of part (a) and part (b) realizes the detailed description of information such as vehicle outline, and makes the present invention adapt to various weather conditions at the same time.

(3) In vehicle detection, the detection process of part (a) and part (b) is independent, and the vehicle detection is realized by combining the candidates of part (a) and part (b), so that the vehicle model adapts to the actual vehicle Moderate deformation of , while reducing the impact of occlusion on vehicle detection.

Description of drawings

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

Fig. 1 is a complex traffic scene with vehicle occlusion in the embodiment of the present invention;

Fig. 2 is the image area corresponding to parts (a) and (b) in the embodiment of the present invention;

FIG. 3 is a schematic diagram of a part of training images in an embodiment of the present invention;

Fig. 4 is the mixed image template corresponding to part (a) and part (b) in the embodiment of the present invention;

Fig. 5 is a vehicle detection result on a test traffic image in an embodiment of the present invention.

Detailed ways

As shown in Figures 1-5, the implementation of the present invention is divided into four main steps: part selection, part modeling, model learning and vehicle detection. These four steps are described in detail below:

Step S1: In part selection, considering the vehicle occlusion in complex traffic scenes (as shown in Figure 1), select two parts from the vehicle object according to the degree of easy occlusion to form the vehicle model. These two parts are parts ( a) and part (b), wherein part (a) is an area that is not easily blocked near the car window; part (b) is an area that is easily blocked around the license plate. In the embodiment of the present invention, part (a) includes the whole car window, a part of the roof near the car window and a part of the hood near the car window and other parts where the image area is located and the image area around these parts; part (b) includes the car Parts such as lights, license plates, and a portion of the hood near the lights, and the image area around these parts.

Step S2: In part modeling, the present invention uses a mixed image template to model the part (a) and part (b). A mixed image template contains one or more image blocks. The image blocks are divided into: edge block, texture block, color block and flatness block. The image blocks in the mixed image template can be one or more of these types .

The edge blocks are represented by Gabor wavelet primitives of a specific direction. The embodiment of the present invention uses Gabor wavelet primitives with 16 directions as an example to represent different edge blocks. Of course, it is only necessary to select Gabor wavelet primitives with no less than one direction here, not limited to 16 directions.

The texture blocks are represented by gradient histograms of image regions. In the embodiment of the present invention, the gradient histogram is obtained by counting the Gabor filter response values in 16 directions of the image area. Of course, it is only necessary to count the Gabor filter response values in no less than one direction, and it is not limited to 16 direction.

The color patches are represented by a color histogram of the image region. In the embodiment of the present invention, the color histogram is obtained by counting the pixel values of three color channels in the HSV color space of the image area. Of course, it is not limited to the HSV color space, and is not limited to three color channels, as long as there are many Just use 1 color channel.

The flatness block is represented by a value obtained by superimposing Gabor filter response values in one or more directions in the image region. In the embodiment of the present invention, the flatness block is represented by the value obtained by superimposing the Gabor filter response values in 16 directions in the image area. Of course, it is only necessary to superimpose the Gabor filter response values in no less than 1 direction, and it is not limited to 16 direction.

Step S3: Model learning includes the following three steps:

Step S3-1, intercepting vehicle images from actual traffic images as training images, the embodiment of the present invention uses 20 training images, of course, as long as no less than 1 training image is used here, it is not limited to 20 training images, And the more the merrier. Figure 3 shows a schematic diagram of a part of the training images.

Step S3-2, using the information mapping method (Information Projection Principle) to learn the image blocks in the mixed image template of the part (a) and part (b) from the training image, and at the same time obtain the part (a) or Probability distribution of mixed image templates for part (b)

$\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; \&Lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; q</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; \&Pi;</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}\frac{\mathrm{<span\; class="notranslate">\; exp</span>}<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; \&Lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span>}{{\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

是p_i中第j个图像块和图像

之间的距离，Z_ij是归一化常数。图4展示了本发明实施例中学习出的所述部分(a)和部分(b)的混合图像模板。where p ₁ and p ₂ are the mixed image templates of the part (a) and part (b) respectively, N _i is the number of image patches in _pi , q(I) is a reference distribution, λ _ij is _the The coefficient of the jth image block,

is the jth image block and image in p _i

The distance between Z _ij is the normalization constant. Fig. 4 shows the mixed image templates of the part (a) and part (b) learned in the embodiment of the present invention.

Step S3-3, learning the probability distribution of the position and scale relationship between the part (a) and the part (b) from the training image. In the embodiment of the present invention, the probability distribution of the position and scale relationship between part (a) and part (b) is set to obey the Gaussian distribution, and the parameters of the Gaussian distribution are learned from the training images. Of course, according to the actual situation, the probability distribution of the position and scale relationship between part (a) and part (b) can also obey other types of distribution, not limited to Gaussian distribution.

Step S4: The vehicle detection process is an iterative process, including:

Step S4-1, using the mixed image template of the part (a) and part (b) to perform template matching on the test traffic image, and extract candidates for the part (a) and part (b). In the embodiment of the present invention, when performing template matching on the test traffic image, the test traffic image will be scaled to adapt to the size of the mixed image template of part (a) and part (b), so as to achieve the purpose of scale transformation.

Step S4-2, using the position and scale relationship of the part (a) and part (b) to combine part candidates to generate one or more vehicle candidates, and calculating the vehicle detection probabilities corresponding to these vehicle candidates. The vehicle detection probability is:

$\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \}</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; \&Lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; \&Lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; \&Pi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{{\mathrm{<span\; class="notranslate">\; \&Lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

和

为在车辆检测过程中被检测出的所述部分(a)和(b)的候选者所在的图像区域，表示了一对部分(a)和(b)的候选者之间的位置和尺度关系的概率。Among them, I is a test traffic image,

and

is the image area where the candidates of the parts (a) and (b) are detected during the vehicle detection process, Denotes the probability of the location and scale relationship between a pair of candidates for parts (a) and (b).

Step S4-3, extracting vehicle candidates with maximum vehicle detection probability.

Step S4-4, comparing the maximum vehicle detection probability in the step S4-3 with the vehicle detection threshold. If the maximum vehicle detection probability is greater than or equal to the vehicle detection threshold, the corresponding vehicle candidate is a detected vehicle object, and then the detected vehicle object is erased from the test traffic image, and the test traffic image of the detected vehicle object is erased Perform the next iterative step (ie repeat steps S4-1, S4-2, S4-3, S4-4). If the maximum vehicle detection probability is less than the vehicle detection threshold, the entire iterative process stops and the detection process ends. Fig. 5 shows the vehicle detection results of the present invention in a complex traffic scene.

In addition, the calculation process of the vehicle detection threshold includes:

First, using the vehicle detection method in step S4 to detect vehicles from all the training images, and calculate the corresponding vehicle detection probability;

Then, the vehicle detection threshold is estimated using the vehicle detection probabilities of all the training images.

The above is a description of specific implementations of the present invention, and is not intended to limit the protection scope of the present invention; all equivalent solutions that can be obtained according to the foregoing descriptions shall be included in the protection scope of the present invention.

Claims (15)

1. the vehicle checking method based on department pattern, is characterized in that: mainly comprise part selection, part modeling, model learning and vehicle detection;

Described part is selected, according to the degree of easily blocking Extraction parts (a) and part (b) two parts from Vehicle Object, for forming auto model; Partly (a) is difficult for the zone be blocked near vehicle window, and partly (b) is the zone easily be blocked around car plate;

Described part modeling, utilize and merge described two parts of manifold vision-mix template modeling;

Described model learning, utilize training image to learn the vision-mix template of described two parts and position and the scaling relation between two parts;

Described vehicle detection, utilize iterative process to detect the one or more vehicles in the test traffic image, in each iterative step, at first utilize template matches to detect two-part candidate, then utilize position and scaling relation built-up section candidate between two parts, realize vehicle detection.

2. vehicle checking method according to claim 1, it is characterized in that: described model learning comprises the following steps:

Step S3-1, from the actual traffic image, the intercepting vehicle image is as training image, and training image quantity is no less than 1 width;

Step S3-2, utilize all image blocks of message maps method from the vision-mix template of described training image learning part (a) and part (b) correspondence and the image likelihood probability of part (a) and the vision-mix template that partly (b) is corresponding;

Step S3-3, utilize described training image study part (a) and the partly position between (b) and the probability distribution of scaling relation.

3. vehicle checking method according to claim 1 is characterized in that: described vehicle detection, and utilize an iterative process to detect one or more vehicles from the test traffic image, comprise the following steps:

Step S4-1, utilize the vision-mix template of described part (a) and part (b) to carry out template matches to the test traffic image, extracts the candidate of described part (a) and part (b);

Step S4-2, utilize position and the scaling relation of described part (a) and part (b) that the candidate of part (a) and part (b) is combined out to one or more vehicle candidates, and calculate the vehicle detection probability that these vehicle candidates are corresponding;

Step S4-3, extract the vehicle candidate with maximum vehicle detection probability;

Step S4-4, compare maximum vehicle detection probability and vehicle detection threshold value in described step S4-3, if this vehicle detection probability is more than or equal to the vehicle detection threshold value, corresponding vehicle candidate is a detected Vehicle Object; Then utilize the test traffic image of erasing this detected Vehicle Object to carry out next iterative step, repeating step S4-1, S4-2, S4-3, S4-4; If this vehicle detection probability is less than the vehicle detection threshold value, whole iterative process stops, and the vehicle detection process finishes.

4. vehicle checking method according to claim 2 is characterized in that: described vehicle detection, and utilize an iterative process to detect one or more vehicles from the test traffic image, comprise the following steps:

Step S4-1, utilize the vision-mix template of described part (a) and part (b) to carry out template matches to the test traffic image, extracts the candidate of described part (a) and part (b);

Step S4-2, utilize position and the scaling relation of described part (a) and part (b) that the candidate of part (a) and part (b) is combined out to one or more vehicle candidates, and calculate the vehicle detection probability that these vehicle candidates are corresponding;

Step S4-3, extract the vehicle candidate with maximum vehicle detection probability;

Step S4-4, compare maximum vehicle detection probability and vehicle detection threshold value in described step S4-3, if this vehicle detection probability is more than or equal to the vehicle detection threshold value, corresponding vehicle candidate is a detected Vehicle Object; Then utilize the test traffic image of erasing this detected Vehicle Object to carry out next iterative step, repeating step S4-1, S4-2, S4-3, S4-4; If this vehicle detection probability is less than the vehicle detection threshold value, whole iterative process stops, and the vehicle detection process finishes.

5. according to the described vehicle checking method of claim 1 to 4 any one, it is characterized in that: described vision-mix template is comprised of one or more image blocks, image block is divided into the types such as edge block, texture block, color block and flatness piece, and the image block that described vision-mix template comprises can be one or more in these types;

Described edge block is that the Gabor small echo primitive by specific direction means; Described texture block is that the histogram of gradients by image-region means; Described color block is meaned by the color histogram of image-region; The value representation that described flatness piece is obtained by the response of the Gabor wave filter of one or more directions in the superimposed image zone.

6. according to claim 2,3 or 4 described vehicle checking methods, it is characterized in that: the image likelihood probability of the vision-mix template of described part (a) or part (b) is:

$p\left(I_{{\mathrm{\&Lambda;}}_i}\right|p_i)=q\left(I\right){\mathrm{\&Pi;}}_{j=1}^{N_i}\frac{\exp \left\{{\mathrm{\&lambda;}}_{\mathrm{ij}}f\left(I_{{\mathrm{\&Lambda;}}_{\mathrm{ij}}}\right)\right\}}{Z_{\mathrm{ij}}},$

P wherein ₁and p ₂be respectively the vision-mix template of described part (a) and part (b), N _ip _ithe quantity of middle image block, q (I) is a reference distribution, λ _ijp _iin the coefficient of j image block,

p _iin j image block and image

between distance, Z _ijit is normaliztion constant.

7. vehicle checking method according to claim 5 is characterized in that: the image likelihood probability of the vision-mix template of described part (a) or part (b) is:

$p\left(I_{{\mathrm{\&Lambda;}}_i}\right|p_i)=q\left(I\right){\mathrm{\&Pi;}}_{j=1}^{N_i}\frac{\exp \left\{{\mathrm{\&lambda;}}_{\mathrm{ij}}f\left(I_{{\mathrm{\&Lambda;}}_{\mathrm{ij}}}\right)\right\}}{Z_{\mathrm{ij}}},$

P wherein ₁and p ₂be respectively the vision-mix template of described part (a) and part (b), N _ip _ithe quantity of middle image block, q (I) is a reference distribution, λ _ijp _iin the coefficient of j image block,

p _iin j image block and image between distance, Z _ijit is normaliztion constant.

8. according to the described vehicle checking method of claim 3 or 4, it is characterized in that: described vehicle detection probability is:

$p\left(I\right|\{p_1,p_2\})=p(I_{{\mathrm{\&Lambda;}}_1},I_{{\mathrm{\&Lambda;}}_2}){\mathrm{\&Pi;}}_{i=1}^{2}p\left(I_{{\mathrm{\&Lambda;}}_i}\right|p_i)$

Wherein I is a width test traffic image, with

be respectively the image-region at the candidate place of the described part (a) that is detected and part (b) in the vehicle detection process,

meaned a pair of part (a) and the part (b) candidate between position and the probability of scaling relation.

9. vehicle checking method according to claim 5, it is characterized in that: described vehicle detection probability is:

$p\left(I\right|\{p_1,p_2\})=p(I_{{\mathrm{\&Lambda;}}_1},I_{{\mathrm{\&Lambda;}}_2}){\mathrm{\&Pi;}}_{i=1}^{2}p\left(I_{{\mathrm{\&Lambda;}}_i}\right|p_i)$

Wherein I is a width test traffic image,

with

be respectively the image-region at the candidate place of the described part (a) that is detected and part (b) in the vehicle detection process,

meaned a pair of part (a) and the part (b) candidate between position and the probability of scaling relation.

10. vehicle checking method according to claim 6, it is characterized in that: described vehicle detection probability is:

$p\left(I\right|\{p_1,p_2\})=p(I_{{\mathrm{\&Lambda;}}_1},I_{{\mathrm{\&Lambda;}}_2}){\mathrm{\&Pi;}}_{i=1}^{2}p\left(I_{{\mathrm{\&Lambda;}}_i}\right|p_i)$

Wherein I is a width test traffic image, with

be respectively the image-region at the candidate place of the described part (a) that is detected and part (b) in the vehicle detection process,

meaned a pair of part (a) and the part (b) candidate between position and the probability of scaling relation.

11. vehicle checking method according to claim 7 is characterized in that: described vehicle detection probability is:

$p\left(I\right|\{p_1,p_2\})=p(I_{{\mathrm{\&Lambda;}}_1},I_{{\mathrm{\&Lambda;}}_2}){\mathrm{\&Pi;}}_{i=1}^{2}p\left(I_{{\mathrm{\&Lambda;}}_i}\right|p_i)$

Wherein I is a width test traffic image,

with be respectively the image-region at the candidate place of the described part (a) that is detected and part (b) in the vehicle detection process,

meaned a pair of part (a) and the part (b) candidate between position and the probability of scaling relation.

12. according to the described vehicle checking method of claim 3 or 4, it is characterized in that: the calculation procedure of described vehicle detection threshold value comprises:

At first, utilize step S4-1, S4-2, S4-3 detects vehicle from all described training images, and calculates corresponding vehicle detection probability;

Then, utilize the vehicle detection probability estimate vehicle detection threshold value of all described training images.

13. vehicle checking method according to claim 5 is characterized in that: the calculation procedure of described vehicle detection threshold value comprises:

At first, utilize step S4-1, S4-2, S4-3 detects vehicle from all described training images, and calculates corresponding vehicle detection probability;

Then, utilize the vehicle detection probability estimate vehicle detection threshold value of all described training images.

14. vehicle checking method according to claim 7 is characterized in that: the calculation procedure of described vehicle detection threshold value comprises:

At first, utilize step S4-1, S4-2, S4-3 detects vehicle from all described training images, and calculates corresponding vehicle detection probability;

Then, utilize the vehicle detection probability estimate vehicle detection threshold value of all described training images.

15. vehicle checking method according to claim 11 is characterized in that: the calculation procedure of described vehicle detection threshold value comprises:

At first, utilize step S4-1, S4-2, S4-3 detects vehicle from all described training images, and calculates corresponding vehicle detection probability;

Then, utilize the vehicle detection probability estimate vehicle detection threshold value of all described training images.

Images