Human tracking in dynamic scenes has been an important topic of research. Objects in the world exhibit complex interactions when captured in a video sequence, some interactions manifest themselves as occlusions. A visual tracking system must be able to track objects which are partially or even fully occluded. In this paper, we propose an approach for tracking multiple objects in dynamic scenes to handle objects partial occlusion. The algorithm consists of two steps at first step we use Gaussian Mixture Model as an effective way to extract moving objects from a video sequence. Then a combinational method is used for shadow removal. The second step of the proposed algorithm is object tracking framework based on Kalman filtering which uses Stable Marriage Problem (SMP) implemented algorithm which is adapted to perform data association and occlusion detection, and fast mean shift computation results for tracking during occlusion. Our approach has the advantages of low cost and low complexity, and can be realized in real time system and is tested on using several real world datasets. © 2012 IEEE.