A Video-Based Traffic Congestion Monitoring System Using Adaptive Background Subtraction

The importance of effective and efficient traffic congestion monitoring grows with the enlarging of urban scale and increasing number of vehicles. We develop a traffic congestion monitoring system which is based on adaptive background subtraction. The system reads real time monitoring video from communications department and converts it into images. After that, we change them into corresponding gray images and carry out image binarization with dynamic multiple thresholds method which selects thresholds depending on pixel, grayscale and pixel position. Afterwards we perform noise reduction with an adaptive median filtering which, taking environmental and other factors into account, dynamically changes median filtering window scale in accordance with the noise density. To fit actual environmental changing, the system updates the background periodically by dynamic background refreshing method. We also put forward an adaptive background subtraction method, which can remove burst noise, to identify the moving objects and get total movement in a given time. Finally, the system determines whether the congestion occurs by comparison result of the total movement and predefined threshold. With the system, traffic management department can facilitate rapid access to the road traffic conditions and real-time traffic congestion monitoring.