The Research And Realization Of Vehicle Flow Detection System Based On Video

In recent years, with the massive growth in the number of motor vehicles brought distress, how to improve the traffic situation become a social core problem to be solved. In this context, intelligent transportation systems(ITS) have come into being.Significance of the study is to protect travel safety of the people, improve the efficiency of the transport system, to improve the living environment and save social resources.The research of vehicle flow detection based on video is the core of intelligent transportation systems, although there has been some significant research results, but there are still some difficult issues to be resolved and improved. For example, in the case of fog, the noise present in the entire sequence of images, the vehicle object detection has serious interference; as the vehicle images exist shadow, both the vehicle itself and the vehicle shadows have the same speed and direction of movement,often in the detection of process, which would be classified as the vehicle object.Above is an objective reality and urgent problem.This article mainly aims at the above two cases are discussed and the research,main content and innovation points are as follows:(1)In view of the the traditional vehicle object detection methods have the lower accuracy under normal circumstances and foggy environment testing, we propose a vehicle flow detection algorithm based on low-rank matrix.The algorithm firstly introduce the Ising model and Robust Principal Component Analysis to get the no-convex energy function, and then employ the singular value decomposition and iterate to solve the problem that energy function is non-convex, and then optimize the energy function to detect the foreground vehicles. Compared with the frame-difference method and the mixed Gaussian algorithm, the experimental results show that the proposed method can detect vehicle effectively and accurately, even in fog weather conditions can also be good segment the vehicle object.(2)In view of the the traditional vehicle object detection methods have the lower accuracy in the shadow of the scene,proposed a large regions texture-based shadow detection method in the HSV color space.This method proceed from the departure from the shadows and background has a similar texture features, the first use intensityfeature and the chromacity feature separated from HSV color space to search candidate shadow pixels, then depent on the pixel correlation of the candidate shadow pixels are combined into a large candidate shadow region. Finally, according to polar coordinates to obtain the pixel magnitude and the pixel gradient direction,the edge pixels of the candidate shadow pixels as the significant pixels which have a large magnitude pixel,to obtain the final shadow area by calculating the significant pixels gradient direction correlation.Experimental results show that this method can be a good deal with the problem of vehicle detection in the presence of shadows, and has better robustness.