Multiple people tracking using moment based approach

Abstract

This paper has the capability to detect multiple people in indoor and outdoor environment. In this paper we have used single camera. In this paper we proposed a technique in which it performs multiple face detection, from this it extracts the people's torso regions and stores the HSV range of each person. After this when person's face is not in front of the camera it will track all those people's using moment based approach i.e. it will compute the area of exposed torso region and centre of gravity of the segmented torso region of each person. In this paper we consider torso region HSV range as the key feature. From this we calculate the tracking parameters for each person. In this paper speech thread module is implemented to have interaction with the system. Experiment results validate the robust performance of the proposed approach. © Springer International Publishing Switzerland 2014.