Research On Moving Object Detection And Tracking Algorithm Based On Vehicle-loaded Video

Vehicle detection and tracking algorithm is the core of Safety Assistant Driving System, also the hot research area of intelligent vehicle navigation, real-time, accurate and robust detection and tracking algorithm can make intelligent vehicle implement the effective early-warning and control of the deviation prevention and anti-collision conditions. The research target of the article is to analyze video image information captured by vehicle camera through the image processing algorithm, detect and track preceding moving vehicle, provide real-time and accurate road information for the driver, and avoid traffic accidents.The paper described the research background and significance of vehicle detection and tracking algorithm, studied and analyzed the basic method of image preprocessing, lane recognition, vehicle detection and vehicle tracking algorithm. Based on the step change of gray of the sky and road intersection, a method of utilizing horizontal gray value distribution curve for extracting road region was given; the lane was detected through combining Canny algorithm with Hough transform, and tracked through more advanced linear-parabolic model; the determination of vehicle hypothesis region was based on the shadow underneath a vehicle and pinhole imaging principle, on the basis of the study of multi-feature fusion algorithm to verify the region of interest, a method of utilizing texture, vertical gradient projection mean and edge symmetry to verify the hypothesis region hierarchically was given; vehicle tracking algorithm based on Kalman filtering and its advantage and disadvantage were analyzed furtherly, considering problems such as divergence, the difficulty of selecting the initial value, and the hypothesis of noise parameter and motion state, preceding vehicle tracking algorithm was given on the basis of gray model GM(1,1), which was modified through introducing the prediction error, and was able to track vehicle real-timely and accurately. Software design of preceding vehicle detection and tracking system was completed, single frame detection and real-time tracking function were also realized. In this paper, we used PETS2001 image sequence to experiment and verified the validity of the algorithm from real-time and accuracy two aspects, also had a clear direction of future research and potential applications.In this paper, vehicle detection and tracking technology is explored and researched, which has some reference value in Safety Assistant Driving System.