CN104658254A - Motorcycle detection method for traffic videos - Google Patents

Abstract

The invention relates to a motorcycle detection method for traffic videos. According to the method, a plurality of detection frames are acquired in a window screening manner. the method specifically comprises steps as follows: 1) extracting the HoG feature from each detection frame to acquire the HoG confidence; 2) extracting the LBP feature from each detection frame to acquire the LBP confidence; 3) extracting the color feature from each detection frame to acquire the color confidence; 4) obtaining motorcycle detection results according to the HoG confidence, the LBP confidence and the color confidence acquired in Steps 1)-3) on the basis of an SVM classifier; 5) judging whether moving speeds of the detection frames corresponding to motorcycles are within the threshold range or not, if the moving speeds of the detection frames corresponding to the motorcycles are within the threshold range, determining that detection results are correct, otherwise, determining that misinformation occurs. Compared with the prior art, the method has the advantages of low misinformation probability, simplicity in implementation and the like.

Description

A kind of motorcycle detection method of traffic video

Technical field

The present invention relates to traffic video detection field, especially relate to a kind of motorcycle detection method of traffic video.

Background technology

Every year because motorcycle causes traffic accidents kill very many; because its speed is fast, poor performance, safeguard measure are weak, very easily there is traffic hazard in motorcycle, and accident occur after injures and deaths extremely serious; mostly cause head injuries, the main cause that this is motorcycle accident mortality ratio, disability rate is high.Capture in traffic video motorcycle, electric motor car, bicycle to traffic violations detect have a lot of help, such as can judge whether to travel on car lane.Conventional motorcycle detection algorithm mainly have employed the mode that HOG and SVM combines.The major advantage of the method be speed fast but when practical application performance do not satisfy the demands, as there is the situation such as more wrong report, Some vehicles loss.

Summary of the invention

Object of the present invention be exactly in order to overcome above-mentioned prior art exist defect and the motorcycle detection method that a kind of misinformation probability is low, implement simple traffic video is provided.

Object of the present invention can be achieved through the following technical solutions:

A motorcycle detection method for traffic video, the method adopts the mode based on window scanning to obtain multiple detection block, specifically comprises the following steps:

1) HoG feature is extracted to each detection block, obtain HoG degree of confidence;

2) LBP feature is extracted to each detection block, obtain LBP degree of confidence;

3) color characteristic is extracted to each detection block, obtain color degree of confidence;

4) based on SVM classifier, according to step 1)-3) the HoG degree of confidence that obtains, LBP degree of confidence, color degree of confidence obtain motorcycle testing result;

5) judge that the translational speed of the detection block corresponding to motorcycle is whether in threshold range, if so, then judge that testing result is correct, if not, be then judged to be wrong report.

Described step 1) and step 2) in, use SVM classifier to classify to HoG feature, LBP feature respectively, and then obtain HoG degree of confidence and LBP degree of confidence respectively.

Described step 3) be specially:

301) for a detection block, get the rectangle frame of any two positions in this detection block respectively, calculate the similarity of the color histogram of two rectangle frames, and preserve;

302) step 301 is repeated) D time, obtain a D group color feature vector;

303) step 301 is repeated) and 302), until the color feature vector of all detection block extracts complete;

304) adopt AdaBoost algorithm to classify, obtain color degree of confidence.

Described step 302) in, the value of number of times D is 100K ~ 1M.

Described step 304) in, provide by the multiple Weak Classifiers in AdaBoost sorter the judged result whether detection block is pedestrian, the judged result of multiple Weak Classifier is weighted on average, obtains final AdaBoost algorithm classification result.

Described color histogram adopts hsv color space, and wherein each Color Channel is divided into K interval.

The value of described K is 6.

Described step 5) in, threshold range is obtained by the labeled data at crossing.

The method also comprises:

Adopt texton boost Algorithm for Training to obtain road surface sorter, be that the detection block of motorcycle inputs in the sorter of described road surface by testing result, judge in this detection block, whether motorcycle lower zone is road surface, if, then testing result is correct, if not, then and testing result mistake.

Compared with prior art, the present invention has the following advantages:

(1) the HoG degree of confidence of each detection block, LBP degree of confidence, color degree of confidence consider by the present invention, and motorcycle testing result precision is high;

(2) the present invention combines the translational speed of motorcycle when detecting, and decreases wrong report;

(3) the present invention also improves accuracy of detection further by road surface sorter, effectively reduces misinformation probability;

(4) the present invention program is simple, easy to implement;

(5) the inventive method is applicable to the detection of motorcycle, electric motor car, bicycle, applied widely.

Accompanying drawing explanation

Fig. 1 is schematic flow sheet of the present invention.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.The present embodiment is implemented premised on technical solution of the present invention, give detailed embodiment and concrete operating process, but protection scope of the present invention is not limited to following embodiment.

Embodiment 1

As shown in Figure 1, the present embodiment also provides a kind of motorcycle detection method of traffic video, and the method adopts the mode based on window scanning to obtain multiple detection block, can be applicable to the detection of motorcycle, electric motor car, bicycle simultaneously.The method specifically comprises the following steps:

Step S1, extracts HoG feature to each detection block, uses SVM classifier to classify, obtains HoG degree of confidence.

Step S2, extracts LBP feature to each detection block, uses SVM classifier to classify, obtains LBP degree of confidence.

Step S1 and step S2 is prior art conventional means.

Step S3, extracts color characteristic to each detection block, obtains color degree of confidence.The car sound color of motorcycle and the clothes color of motorcyclist pedestrian have similarity: can have a variety of color, but in most cases, the color of self is similar.According to this information, propose the feature of color self-similarity.Be specially:

301) for a detection block, if size is 32*64, get the rectangle frame of any two positions in this detection block, if size is 8*8, calculate the similarity of the color histogram of two rectangle frames, and preserve;

Described color histogram adopts hsv color space, and wherein each Color Channel is divided into K interval, and the value of K is 6.

The computing formula of the similarity of the color histogram of two rectangle frames is:

$C(p_i,q_j)=\frac{1}{2}\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\frac{{[h_i\left(k\right)-h_j\left(k\right)]}^2}{h_i\left(k\right)+h_j\left(k\right)}$

P _iand q _jrepresent the rectangle frame of two positions respectively, h _i(k) and h _jk () is p respectively _iand q _jcolor histogram, k is histogram number, and each color road is divided into 6 intervals, always has 216 intervals.

302) step 301 is repeated) D time, obtain a D group color feature vector.Select optimum feature because AdaBoost has, therefore the value of D is the bigger the better, and can cover various situation like this.For the consideration of training speed, the span of final D is between 100K ~ 1M.

303) step 301 is repeated) and 302), until the color feature vector of all detection block extracts complete.

304) adopt AdaBoost algorithm to classify, obtain color degree of confidence.Provide by the multiple Weak Classifiers in AdaBoost sorter the judged result whether detection block is motorcycle, the judged result of multiple Weak Classifier is weighted on average, obtains final AdaBoost algorithm classification result.

Step S4, based on SVM classifier, according to HoG degree of confidence, LBP degree of confidence, color degree of confidence acquisition motorcycle testing result that step S1-S3 obtains.

Step S5, judges that the translational speed of the detection block corresponding to motorcycle is whether in threshold range, if so, then judges that testing result is correct, if not, be then judged to be wrong report.Threshold range is obtained by the labeled data at crossing, i.e. the translational speed scope of motorcycle on picture.

Embodiment 2

Shown in figure 1, in order to further improve accuracy of detection, the detection method that the present embodiment provides also comprises final step:

Utilize texture and color characteristic training to obtain road surface sorter, the present embodiment adopts texton boost algorithm, and it is the region on road surface that algorithm is finally understood in output image; Be that the detection block of motorcycle inputs in the sorter of described road surface by testing result, judge in this detection block, whether motorcycle lower zone is road surface, if so, then testing result is correct, if not, then and testing result mistake.

All the other are with embodiment 1.

Claims (9)

1. a motorcycle detection method for traffic video, is characterized in that, the method adopts the mode based on window scanning to obtain multiple detection block, specifically comprises the following steps:

1) HoG feature is extracted to each detection block, obtain HoG degree of confidence;

2) LBP feature is extracted to each detection block, obtain LBP degree of confidence;

3) color characteristic is extracted to each detection block, obtain color degree of confidence;

4) based on SVM classifier, according to step 1)-3) the HoG degree of confidence that obtains, LBP degree of confidence, color degree of confidence obtain motorcycle testing result;

5) judge that the translational speed of the detection block corresponding to motorcycle is whether in threshold range, if so, then judge that testing result is correct, if not, be then judged to be wrong report.

2. the motorcycle detection method of traffic video according to claim 1, it is characterized in that, described step 1) and step 2) in, use SVM classifier to classify to HoG feature, LBP feature respectively, and then obtain HoG degree of confidence and LBP degree of confidence respectively.

3. the motorcycle detection method of traffic video according to claim 1, is characterized in that, described step 3) be specially:

301) for a detection block, get the rectangle frame of any two positions in this detection block respectively, calculate the similarity of the color histogram of two rectangle frames, and preserve;

302) step 301 is repeated) D time, obtain a D group color feature vector;

303) step 301 is repeated) and 302), until the color feature vector of all detection block extracts complete;

304) adopt AdaBoost algorithm to classify, obtain color degree of confidence.

4. the motorcycle detection method of traffic video according to claim 3, is characterized in that, described step 302) in, the value of number of times D is 100K ~ 1M.

5. the motorcycle detection method of traffic video according to claim 3, it is characterized in that, described step 304) in, the judged result whether detection block is pedestrian is provided by the multiple Weak Classifiers in AdaBoost sorter, the judged result of multiple Weak Classifier is weighted on average, obtains final AdaBoost algorithm classification result.

6. the motorcycle detection method of traffic video according to claim 3, is characterized in that, described color histogram adopts hsv color space, and wherein each Color Channel is divided into K interval.

7. the motorcycle detection method of traffic video according to claim 6, is characterized in that, the value of described K is 6.

8. the motorcycle detection method of traffic video according to claim 1, is characterized in that, described step 5) in, threshold range is obtained by the labeled data at crossing.

9. the motorcycle detection method of traffic video according to claim 1, it is characterized in that, the method also comprises:

Adopt texton boost Algorithm for Training to obtain road surface sorter, be that the detection block of motorcycle inputs in the sorter of described road surface by testing result, judge in this detection block, whether motorcycle lower zone is road surface, if, then testing result is correct, if not, then and testing result mistake.