MULTI-PEOPLE TRACKING ACROSS MULTIPLE CAMERAS

Multi-target tracking (MTT) is an active and challenging research topic. Many different approaches to MTT problem exist, yet there are still few satisfactory methods of solving multi-target occlusion problem, which often appears in multi-target tracking task. The application of multi cameras in most existing researches for multi-target occlusion requires camera calibration parameters in advance, which is not practical in the case of outdoor video surveillance. Most of the proposed solutions for this problem require camera calibration parameters that make them impractical for outdoor video surveillance applications. To address this problem we propose in this paper a probabilistic approach, the foremost consideration of which is to reduce the dependency on camera calibration for multiple camera collaboration. More robustness on target representation and object tracking has been ensured by combining multiple cues such as border information of the object with color histogram, while Gale-Shapley algorithm (GSA) has been used for finding the stable matching between objects of two or more camera views. Efficient tracking of object ensures proficient recognition of target depicting parameters (i.e. apparent color, height and width information of the object) as a consequence provides better camera collaboration. Initial simulation results prove the validity of the proposed approach.