Study Of The Traffic Incident Detection Algorithm Based On Video

Video-based traffic incident detection system can detect the broken down vehicles,retrograde vehicles,thrown objects and other abnormal incidents on the road real-time,which plays an important role in avoiding secondary accidents,it's a research focus at home and abroad currently and have broad application prospects.This paper studies the traffic incident detection lgorithms,the main contents are as follows:(1) Background Estimation. On the basic of analyzing the traditional background estimation methods,a new background estimation method is proposed.This method divide a frame of image into a number of blocks with fixed size. Calculating the mean and the variance of each block in the time series images,by Statistics the probability of the mean and the variance fall into each area in the two-dimensional palne of which consists the mean and the variance ,the background image is estimated in blocks.(2) Image Binarization Segmentation. On the basic of analyzing the traditional image binarization methods,a new image binarization method is proposed.This method divide a frame of image into a number of blocks with fixed size.By changing every pixel gray value in each block in the froground and background images base on the least square method,improve the distribution of the pixel gray balance in a block,and then subtract the pixel in a every block in the forground images from the corresponding pixel in the background images,calculating the sum of the absolute value of the difference in each block,and the image binarization segmentation completed in blocks.(3) Incident Detection. On the basic of analyzing the traditional incident detection methods,a new Incident Detection method is proposed.This method detect the abnormal events bases on cumulative value distribution Of the targeted blocks in the time series.The designed algorithm in this paper experimented in different traffic flow scenarios shows that the new proposed incident algorithm meet the real time detection requirements. The initial application shows that the algorithm has applicability and with good prospects.