Pedestrian Detection and Tracking Based on A Novel Background Subtraction & Particle Filter

Abstract. This paper focuses on pedestrian detection of moving objects, and then tracking the detected objects. The pedestrian detection is performed using an algorithm based on a type of background subtraction technique. This technique not only uses a median-based approach for initial background estimation but also a novel two-pass approach for noise removal. The tracking of the detected object is then done using an offline particle filter. The proposed applications of the algorithm presented in this paper, could be pedestrian activity monitoring, ITS, etc. The experiment indicates that the algorithm can detect and track the pedestrian effectively