A Real-Time Traffic Congestion Detection System Using On-Line Images

The heavily-loaded traffic system in Macao is characterized by narrow and complex street networks, along with many traffic bottlenecks. In this paper, we propose an economical real-time traffic congestion detection system using on-line images provided by the local government. The proposed system mainly consists of the detection of vehicles using the on-line images and the estimation of traffic congestion based on the estimated number of vehicles. For the detection of vehicles, we study a method of using the signs on the road and experiment the technique of using the Haar-like features. We find that Haar-like features can be used for the detection of vehicles using the on-line images from different camera locations. For the traffic congestion estimation, a threshold for the image correlation coefficient of the consecutive images is used, along with a threshold for the number of vehicles detected. Two different levels of congestion are considered, namely NORMAL and CONGESTED, although the number of congestion level can be easily extended. Experimental results show that the proposed system can estimate the traffic congestion correctly and in real-time at low cost. Compared with traditional traffic congestion estimation systems, this system provides a more economical solution with potential commercial applications for the local residents and for the tourists in Macao.