Research And Traffic Detection Technology For Video Based

As people’s living standard rise, cars are becoming ubiquitous in many households. The increasing number of vehicle has brought challenges for the management of traffic. Intelligent transportation system is considered to be an integrated management system to ensure the road traffic real-time, accurate and efficient. Traffic information extraction is one of the most basic part of the intelligent transportation system. With the development of digital image processing technology, the detection method of vehicle based on video processing technology has attracted increasing attentions, and become a major method to obtain traffic information. This paper is the research on road traffic conditions and the analysis of traffic information based on video processing to provide basic data for the improvement of the intelligent transportation system. The main work includes:1) In the research of vehicle detection, the background difference method combining gaussian mixture model background modeling and regional background updating was implemented. Initial background was established by gaussian mixture model, then use Ostu method to find adaptive threshold to segment the foreground; finally background updated non foreground regions. In order to eliminate the influence of shadow, a shadow detection algorithm based on LBP texture and brightness was proposed. First, the LBP texture algorithm was used to detect candidate shadow regions. Second, eliminate the non-shadow pixels by computing and comparing the brightness distortion. Finally, the connected domain processing and region growing processing method was used to improve the shadow region. The experimental results show that the vehicle detection algorithm’s real-time performance is much better than the Gaussian Mixture Model method and its detection effect is similar to gaussian mixture model method, much better than multiframe averaging method. In shadow elimination, results show that the algorithm can accurately detect the shadow pixels.2) In the research of vehicle tracking, a vehicle tracking algorithm based on bounding rectangle and Kalman Filter was researched. First, judging and segmenting the bonded vehicles by contour convex hull characteristics. Then, using kalman filter to predict the next location of vehicle, realizing vehicle tracking and getting the traffic flow.3) On the detection and tracking of vehicles at night, this paper proposes an algorithm of adaptive segmentation threshold based on gray level histogram. Regional growth and morphological processing method are used to improve the candidate headlights regions. The traffic flow was given by headlights pairing and tracking. 4) A prototype of vehicle flow detection system was designed and its main function modules were introduced. To get the vehicle flow data, experiments were done. By comparing the results with those obtained by the virtual coil method, it proved that the system has a higher accuracy of vehicle flow detection.