Study Of Vehicle Parking Detection Algorithms Based On Background Subtraction

Vehicle parking event detection is an important aspect of video traffic detection system. It has a broad application prospect. This paper researches on vehicle parking detection algorithm based on background subtraction.The key steps of the parking event detection method include getting reliable background model, updating background in real time and calculating background subtraction for suspected target. And then the vehicle targets can be detected. The main contents are as follow:(1)Background extraction algorithm. By combining statistic histogram algorithm and Statistical median algorithm, a new background extraction algorithm is proposed. This algorithm gets gray median after removing abnormal gray values from video sequence. The grey value around gray median with highest occurrence frequency is regarded as background gray value.(2)Background updating algorithm.The paper designs a steady-state updating algorithm based on the SAD value of block. By combing the steady-state counter of small difference to gain reliable steady-state degree of confidence to update the background.(3) Parking incident detection. The algorithm uses stable state to distinguish the suspicious blocks. The binary segmentation is operated according to the SAD value of corresponding block between the foreground image and background image. The misinformation of packing caused by the cast falling objects, illumination,water reflection,clouds obscured, is effectively overcame by the mark of connected domain and target tracking technology.The parking detection algorithm designed by the paper has been tested in different traffic scenes. The results show that the algorithm has gained ideal initial background, and the result of background updating can not only meet illumination change but also receives the delayed update for the vehicle parking, it expands the generality of background subtraction and meet the needs of the alarming system for packing event. The algorithm for packing event detection meets the needs of real-time detection and reduces the rate of false alarm caused by various reasons. The initial application shows that the algorithm has good applicability and prospect.