Motion tracking of humans under occlusion using Blobs

Abstract

In today's scenario, Video Surveillance plays a major role in building intelligent systems. This involves the phases such as Motion detection, Object classification and Object tracking. Among these, Object tracking is an important task to identify/detect the objects and track its motion correspondingly. After object identification, the location of the objects is crucial to understand the nature of the moving objects. There arises a need for tracking the occluded objects also when multiple objects are under surveillance. In this paper, a new tracking mechanism has been proposed to track the objects under surveillance though they occlude. Initially, background has been modelled with the Adaptive background modelling using GMM (Gaussian Mixture Model) to obtain the foreground as blobs. Later, Objects represented using Contours are integrated with simple particle filters to obtain a new state which would track the object/person effectively. Using the path estimated by particle filters, Occlusion of blobs gets determined based on the interference of their radii lying in that path. Performance of this system has been tested over CAVIAR and User generated data sets and results seem to be promising. © Springer International Publishing Switzerland 2014.