A Robust Visual Object Tracking Approach on a Mobile Device

摘要

In this paper, we present an approach for tracking an object in video captured on a mobile device. We use a colour-based approach. The performance of many of these approaches degrades due to lighting changes and occlusion. To address the issue of lightning changes, our approach makes use of colour histogram that is generated by accumulating histograms derived from target objects imaged under different conditions. A CAMShift tracking algorithm is applied to the back-projected image to track the target object. We have tested our approach by tracking an Emergency Exit sign and the results obtained show that the tracking is robust against lightning changes.