A ROBUST ILLUMINATION INDEPENDENT OBJECT TRACKING SYSTEM

Abstract

Mobile service providers continuously seek new and innovative ways to personalize information service provided to subscribers. The service that could be of great significance to users is Location Based Information particularly while traveling in an unknown location. This requires identifying and tracking the position of the user. In the current work we establish the location information from an image of the environment. Local image features have shown good results in establishing correspondence between images of the same location taken under differing imaging conditions. In the current work a framework is proposed to extend this approach to locate and track the position of an object/person by matching an image of the location to a large set of images with known location information. The framework is robust in identifying similar images which vary in illumination, are obtained in. presence of occlusion, scale transformation and change in viewpoint. The framework incorporates a conflict resolution mechanism to resolve variance in the location information obtained from different matching images by using a secondary descriptor. The map of a location is represented by means of a graph. Tracking of the user is performed by predicting the path followed• by the person/object between successive queries. This is aided by incorporating additional information into edge weights. The goal of tracking is to collect statistics which could be used for commercial advantage