A NEW APPROACH TO OBJECT LOCALIZATION USING BACKGROUND FEATURES

Abstract

This thesis discusses the study of object tracking technique based upon background features/parameters. This concept is based upon the scientific principle of relative localization and the principle of natural compression with the increase in the distance from the location of camera. The second principle has been utilized for depth modelling of the foreground objects and the background information has been used not only to improve the depth modelling but also to learn to better localize. Method shows a definite improvement in the object localization process. Method also works in real time being having no computationally intensive algorithm.