Research On Video - Based Traffic Statistics Algorithm

In recent years, with the urban scale of Chinese expanding rapidly, our city’s traffic has increased dramatically. Intelligent Transportation System impacting the traffic situation can’t be ignored. It not only has guaranteed the smooth traffic, the safety of pedestrians and vehicles, but also indirectly has contributed to Chinese economic development. In the intelligent transportation system, Vehicle detection technology based on video has become a hotspot. Vehicle detection based on video technology is also used in the traffic statistics, which can obtain total number of vehicles passed on the road in a period, by processing the road traffic video from camera, providing basic data for intelligent traffic management, thereby easing the traffic pressure and improving the rate of road resources utilization.This paper studies algorithms of the traffic statistics system based on video. This paper uses the background subtraction method for moving objects detection. The realization of this algorithm is based on the comprehensive application of image pre-processing, moving target detection, background construction and other algorithms, combined with the shadow detection algorithm of HSV color space eliminating the shadows of moving targets, which let the target detection be more accurate. In addition, the premise of using background subtraction is to extract the ideal background image, this paper proposed a kind of improved background construction algorithm based on statistics, collecting first N frame image pixels, taking pixel values in the highest probability for each pixel as background pixels, thereby to obtain the initialization background images. The algorithm can obtain the ideal background images, however, change of factors such as weather, illumination will affect video background, causing the change of the background images, so it also needs to update the background images, this paper proposed an improved background update algorithm to update the background images. Moving object pixels are replaced with corresponding background image pixels, then by counting to get average to update the background. Finally, this paper uses a combination algorithm of virtual coil and moving object detection for traffic statistics, combined with de-noising, fill holes, edge detection algorithm to determine the vehicle again, in order to prevent the occurrence of missing inspection.This algorithm which was used in this paper has some advantages, such as small amount of calculation, high real-time. The results obtained by experiments shows that the result of traffic statistics obtained by using this integrated algorithm has higher accuracy and feasibility.