MULTIPLE CAMERA BASED OBJECT TRACKING-OBSTACLE DETECTION

Abstract

In this thesis, novel methods for background modelling, tracking and occlusion handling via multi-camera configurations are presented. Specifically, we have developed a system to first track moving persons in a given scene and generate colour-based models of those persons to accomplish identification at a later time. The tracking is non-invasive meaning that it does not require persons to wear any particular electronics or clothing to be able to track them. Tracking is accomplished using a position-based data association algorithm while the colour modelling is accomplished using a mixture-of-Gaussians statistical model. The expectation maximization algorithm is used to generate the colour models over a sequence of frames of data; but to track people successfully in multiple perspective imagery; one needs to establish correspondence between objects captured in multiple cameras. We present a system for tracking people in multiple uncalibrated cameras. The system is able to discover spatial relationships between the camera fields of view (FOV) and use this information to correspond between different perspective views of the same person. We employ the novel approach of finding the limits of field of view (FOV) of a camera as visible in the other cameras. Using this information, when a person is seen in one camera, we are able to predict all the other cameras in which this person will be visible.