Vehicle Flow Detection Algorithm And Implementation Based On Difference Image

Intelligent Transportation System relates to all areas of people's travel. It affects the efficiency of travel. Vehicle flow detection system is a subsystem of intelligent transportation systems. It provides traffic participants and managers with vehicle flow the most basic data. We can extract traffic congestion information, vehicle density, occupancy of road etc. from vehicle flow. Construction and maintenance of traditional suspended systems, such as microwave system, infrared system and radar system are complicated. With the development of information technology, Vehicle flow detection systems based on image processing technology have gradually replaced the traditional traffic detection system. Construction and maintenance of the new system is sample, and detection information is rich.Vehicle flow detection algorithm is the core technology of vehicle flow detection systems based on image processing technology. But there are no universal vehicle flow detection algorithm. Scholars from countries in the world design different vehicle flow detection algorithm according to specific traffic scene. So this paper respectively design vehicle flow detection for highway traffic scene and night traffic video of toll station. The main work and research fruits are as follows:1. Designed mixed difference vehicle flow detection algorithm for highway traffic scene. This algorithm uses background subtraction to detect environment mutation. When there are environment mutations, it calls frame-difference method; otherwise it uses mixed-background subtraction and frame-difference method. Designs virtual detection area setting method and vehicle counting criteria to reduce error from vehicle changing lines and vehicle blocking. The detection results of main lane in the video show that the accuracy rate is 95.45%, undetected rate is 4.55%, both better than background subtraction algorithm. False detection rate of the algorithm is 1.52%, better than the frame difference method which false detection rate is 19.7%.2. Based on the principle of the system availability, night vehicle flow detection algorithm was researched. This paper improved night vehicle flow detection algorithm based on character recognition. This algorithm selected a virtual detecting area to avoid reflection noise from vehicle body and road surface. Introduced the judging method based on area and shape factor to distinguish between vehicle lights and noise from vehicle body and road surface. Adopted a compulsive rejection algorithm to avoid iterative count of vehicles. Test results demonstrate the accuracy is over94.44%. The accuracy increases 3.70% than traditional method. 3. According to structured software designing method, this paper implemented highway mixed difference vehicle flow detection algorithm and improved night vehicle flow detection algorithm, under platform VC++6.0 with OpenCV...