Video Image Processing Algorithms For Traffic Detection

Along with the development of transportation system, the problems caused by traffic safety, congestion, pollution, energy consumption are getting more and more serious. How to improve the level of traffic management and transportation system operation efficiency is in the limelight for the transportation industry over the past decade. As the key technique for the development of transportation technology,traffic flow detection is a hot research topic of intelligent transportation system. This paper focuses on fusion of frame differential method and background subtraction method to develop in-depth research of traffic flow detection system.Firstly, This paper introduced the basic theory of digital image processing:the gray change of images, the balance processing of image, enhancement of gray image and filtering of image. Afterwards, analysis common algorithms of vehicle traffic detection. Comparing with shortcomings of the frame differential method and background subtraction method particularly, it proposes image processing fusion algorithm which based on frame differential method and background subtraction method. The algorithm can blend the advantages of both, and avoid their respectively shortcomings, then make a higher accuracy rate. In addition, on the analysis of the average method of multi frame, block method, etc. commonly used method of adaptability. On the basis of statistics in value method to the modeling of the background to realize the background estimation and background updates. Finally, the paper puts forward an implementation scheme of traffic flow detection system based on video image processing technology, and work out the traffic flow detection of image processing software.The system is designed in the compilation environment of Visual C++, do experiment of traffic flow detection algorithm from the collected video data. Under the conditions of high quality of image, a reasonable distribution of virtual loops, select appropriate threshold, we find that the recognition rate of the algorithm turn to be high and the result achieves a satisfactory effect through experiments and analysis.