Research Of Vehicle Detection And Tracking Based On Traffic Video Sequence

In recent years, with the rapid growth of motor vehicle, urban transportation problem has become more and more serious.Urban transportation problem is universally accepted as a worldwide problem.In our country, for the limitation of road construction lands and funds, it’s extremely urgent to establish complete intelligent transportation system. The image processing and pattern recognition are the core of intelligent transportation system,as to the intelligent transportation system based on the video detection. The paper does research on vehicle detection and tracking algorithm based on the traffic video sequence.On the basis of traditional moving target detection algorithm, and the original SD’s foreground detection and background extraction algorithm, the paper brings in the temporary frame as the basis of background estimation and updating, and puts forward the modified SDTS vehicle detection and background estimation and updating algorithm. By using subjective contrast test and quantitative parameters comparison, the paper proves the instantaneity and accuracy of this algorithm. Then some experiments taken under a variety of climate conditions verify that this method has strong background adaptive capacity and noise rejection ability, and can detect the moving targets effectively. To solve the shadow problem existing in the traffic video, the paper first confirms whether there are shadows. Then a method based on the histogram threshold value and the radio of gray value is used for eliminating the shadows. The paper verifies the effectiveness of the algorithm by the shadow detection testing video.The paper focuses on the edge features tracking algorithm based on Kalman filter, and the region features tracking algorithm based on mean shift vector. Combined with the advantages of both two algorithms, a new vehicle tracking algorithm based on the edge and region features is put forward. After then, the paper verifies its high instantaneity and better tracking effect by the testing video.