Research On Video Vehicle Tracking Method

In comparison with traditional vehicle detector, the vehicle detector based on video image processing and visual technology has many advantages, such as faster processing speed, convenient and fast installation and maintenance at lower cost, wider surveillance range, and more retrievable kinds of traffic parameter and so on. Therefore, it has been more and more widely applied in intelligent traffic system (ITS) recent years. Many video image processing and analyzing technologies have been raised focusing on the video traffic sequence images, in which effective detection and tracking of moving vehicles is the hard-core of modern intelligent traffic system and an important field of computer visual study.This paper detects and tracks moving vehicles from video image sequences extracted from fixed camera, thus extracts traffic parameters to assist in solving traffic problems. Main study tasks finished are as follows:1. Detection of moving objective vehicles. Background subtraction is used to obtain the moving objective. A self-adapting background extraction and updating method based on interval distribution is used to obtain background, and background subtraction combining with shareholding is used to detect moving objective vehicles. The image smoothing, strengthening, boundary detection, noise-wiping-off and so on are analyzed and studied. Effective detection and extraction of moving vehicles from video sequences are realized. To deal with the shadow problems of moving vehicles, the shadow of detective vehicles is wiped off based on existing algorithm.2. Tracking of moving vehicles. An extended-Kalman Filter tracking method based on color features is raised in this paper. This method uses Kalman Filter Theory to build movement model, predicts by means of vehicle features and matches the predicted moving objective with the objective in current frame to determine its motion trail. Track the moving vehicle and acquire important traffic parameters such as vehicle speed, car flow and so on. Since the color feature contains more information than grayscale feature and has translation and rotational invariance, the method can solve the problems of vehicle turning and occlusion properly through extracting the color feature of objective vehicle, and carry on exactly objective location tracking.We simulate the algorithm in the computer and test the performance of traffic detection, shadow removal and vehicle tracking. Experimental results show that the system is more robust and accurate, and look from algorithm realization angle, it possesses the features of easy-to-use and strong real-time property.