Tracking Multiple Objects in Complex Scenes

Automatic video surveillance-based security is one of the most dynamically developing segment of the security industry. There has been a large progress of related vision hardware in the recent years, however, methods used to detect and monitor moving objects still lack the robustness to handle specific events occurring in complex scenes. In this paper we propose an algorithm to resolve occlusion events based on the smoothness constraint of kinematic and color features. The methods are tested using two images sequences and qualitative results are presented in terms of the accuracy and stability of obtained trajectories.