Object Tracking Based on Genetic Algorithm and Kalman Filter

Abstract

Robust object tracking is quite important in computer vision. In this paper, a novel tracking approach for single object which combines genetic algorithm and Kalman filter is proposed. Genetic algorithm is introduced and reasonably applied to find the tracked object in a search area. A further step called multi-blocks voting is exploited for obtaining more accurate object localization. Kalman filter is exploited to both estimate the position of object center and cope with temporary occlusion. Results on real sequences and comparisons with other standard techniques demonstrate the effectiveness of the proposed algorithm.