Body tracking and motion analysis using hybrid model

In this paper we present a new approach for tracking movements of human body in dynamic scenes. The body is characterized by a set of points of interest (PI) issued from Harris corner detector where every one is characterized by its local appearance and its position in each frame of the video sequence. Tracking is achieved by template matching between PI in a pair of consecutive frames in the sequence and is based on various distance calculation criteria. Movement analysis is made by an intelligent decision-making tool based on least squares method, able to estimate direction and orientation of body motion.