Practical application for vision-based traffic monitoring system

This paper presents a practical real-time traffic monitoring system based on object detection and tracking for measuring traffic parameters such as speed and volume. The system presented in this paper is a following work from our previous works in. This monitoring system is currently being used on real traffic environment in Thailand. The results in this paper confirm the validity of the proposed systems in the real-time traffic monitoring system. In the proposed system, background is modeled by using edge information and this model is used for extracting foreground moving objects. The advantage of using edge information to model the background is that it is more robust to the lighting variation. Moreover, it requires less computing resources than intensity based background model, thus; it is suitable for real-time application. The extracted moving objects are then tracked by using Lukas-Kanade (Pyramid) optical flow algorithm. Only the successfully tracked vehicle will be considered for retrieving traffic information. We embedded all the algorithms into a small module, comprising of hardware such as computer and network equipments. These modules are installed at various places along Northern to Southern major roads in Thailand. These modules are running continuously for a year to report traffic data to the server. These data is also released to the public through the website http://traffic.thai.net. The experimental results show that the traffic characteristic obtained from the proposed system is similar to the actual road traffic. It indicates that the system can be used to monitor the traffic condition, which is useful for traffic planning and automatic accident monitoring.