Tracking vehicle targets with large aspect change

Abstract

In this paper we present a novel method for tracking rigid objects (mostly vehicles) in video sequences with cluttered background, obtained from a mobile camera. We estimate the motion model of the target using corners extraction and matching together with LMedS statistical approach. The resultant model assists an active contour (snake) to track the target efficiently. To avoid tracking errors resulting from large aspect change of the target, we propose a snake with an automatic local swelling mechanism. This mechanism enables the snake to include new parts of the target which appear as a result of large aspect change. Several experiments have been conducted to show the promise of our algorithms. Key words: contour tracking , moving target , aspect change , active contour , feature matching