The Design And Implementation Of Vehicle Detection And Traffic Flow Stafistics Platform Based On Video Image Processing

Nowadays, with the high-speed development of social economy and the rapid growing of the number of vehicle, traffic congestion and other problems are becoming more and more serious. In the early 1990s, the research of Intelligent Transportation System(ITS) brought a new opportunity to the transportation field. In this paper, the traffic surveillance video was taken as the research object. The key part of the research work was the vehicle detection and traffic flow statistics algorithms. The algorithms of vehicle detection have been studied and a method based on Markov random field model was proposed. An integrated traffic statistics platform was designed and constructed to realize the intelligent statistics of traffic flow with high accuracy.Using Visual Studio 2008 as the development environment and utilizing OpenCV to process the traffic surveillance video to complete the detection and extraction of the vehicle target. The key contents of this paper are listed as follows:1.Video image pretreatment method for vehicle detection:mainly includes the image grayscale, binarization and de-noising. Particularly in the area of image de-noising, gauss noise and salt-pepper noise were added to the traffic video image respectively. Using the mean filter, median filter and gauss filter to wipe off noise to be ready for the following detection and segmentation of vehicle iarget.2.Vehicle detection algorithm based on Markov random field model:The advantages and disadvantages of several kinds of existing vehicle detection methods have been analyzed. The most popular vehicle detection algorithms based on video image processing were compared such as the frame difference method, background difference method and optical flow method. Then a vehicle detection algorithm based on Markov random field model was proposed. The accuracy of the proposed detection algorithm is higher and the lost rate after segmentation is lower. The proposed algorithm can obtain more vehicle target contour, retain more details and also be able to clearly separate the background and target zone. The problems of low signal-to-noise ratio and weak boundary of the road vehicles images are solved well by the proposed algorithm.3.Design and implement an intelligent traffic flow statistics platform based on OpenCV:Using OpenCV library in Visual Studio 2008 platform to realize traffic statistics. All the software modules, data structure, implementation processes, interfaces, the whole structure and the function of the software were introduced in details in this paper. In the end of the paper, the process and the results of the operation of the detection software were tested and analyzed in detail. Experiments results show that the software can achieve the traffic flow statistics with high accuracy. Both of the miss rate and the false detection rate were lower. It can provide a reliable technological support for urban planning research.