Study On Detection For Moving Vehicles In Video Sequences

Moving objection detection without human interference has been a hot issue which attracts more and more attention in intelligent transportation. Many subjects such as computer vision, the collection of information and the operation of video image are related to this issue. Effective detection and precise tracking of moving vehicle are the hardcore of intelligent traffic research. Corresponding study is done systematically in this paper. Taking the natural road as the background, a kind of moving vehicle auto-detecting and auto-tracking method is adopted. Moreover the software system is designed and implemented. A moving vehicle can be automatically tracked elementarily by using the software system.Vehicle detection based on background subtraction is completed in this paper. Analysis and research are done in image smoothing, enhancement, edge detection and noise elimination. A new detecting labeling method is proposed in the paper by improving the old method. Finally, extensive experiments were performed to evaluate the performances of the above-mentioned methods. Tracking the vehicles through image sequences by the repetition of the "matching-correction-prediction" strategy is adopted. Considering moving distance and area changes of one vehicle in the two consecutive frames are slight, an algorithm tracking of moving vehicles are designed. Two parameters, center and area of moving vehicles, were used to match in the method. The moving model of vehicle is established from Kalman filter technique. Prediction is done according to vehicle features such as central position and areas value. Through the match between perdition object and object of the present frame, the corresponding relations of the moving object in the image series is draw out and moving track is fixed on. According to the tracking results, the important traffic parameters like the vehicle speed, the quantity of the traffic flow and so on can be obtained by post-processing and analysis. It reduces the search region of the matching and decrease the chance of mismatches leading, so as to complete effective vehicle tracing.Because only two parameters were used and the datum was slight, the moving vehicles in the frame were tracked and located the position quickly.A moving objection detecting experimental system which based on video is designed in this paper, and is carried out in Visual C++. The main system frame and implement are given. At last, Analysis about factors that affect the precision of traffic detection is put forward, and some settlements are brought forward.