Pedestrian and bicycle data are necessary for the purpose of transportation planning, infrastructure design and traffic management. Nevertheless such data cannot be directly collected by the commonly used detectors (e.g. inductive loop, sonar and microwave sensors). In this paper, a pedestrian and bicycle tracking and classification (PBTC) system is developed for pedestrian and bicycle detection using video camera. This system contains sixmodules: video flow capture module, moving detection module, shadow removal module, feature extraction module, tracking module, and classification module. Gaussian Mixture Model (GMM) is utilized to extract moving objectsfrom image sequence. In tracking module, the most challenge part of this system, the trajectories are obtained by Kalman filer (KF). To identify pedestrians and bicycles, Back Propagation Neural Network (BPNN) is employedin classification module. Two other simple but effective algorithms are used to alleviate the negative impacts from shadows and occlusion. The system was tested at three test sites under different traffic and environment conditions. It has been confirmed that the accuracy for pedestrian detection was approximate 85% and the count error was lower than 13% for bicycle in all tests sits. This indicates that the proposed system is a feasible alternative for non-motorized data collection.
Abstract