||In this thesis, a human detection and tracking system in a crowded environment is presented. The biggest challenge of the system is to detect occluded people from the captured video, where some visible information of the occluded people in a camera is lost. Many researchers proposed to reconstruct the occluded information from other cameras by fusing the human information captured from several cameras, typically four to eight cameras, with different viewing angles. However, this leads a high computation cost of the system in processing multiple images simultaneously. Besides installing several cameras for reconstructing occluded information, an additional hardware is required to synchronize the captured images from several cameras. This results in a high cost of the hardware platform. Not limited to these, the system cannot provide an easy installation and reconfiguration. In this thesis, a novel occlusion model is proposed to solve the occlusion problem using a stereo camera system. This algorithm is achieved by modeling occlusion events and compensating the loss of visible pixels. This approach significantly reduces the hardware cost and computation cost of the system. Synchronization of multiple cameras is also not required and the setup of the system is much simpler. The result shows that this system is capable of tracking people robustly under severe occlusion and it achieves the best tracking density among most of the similar tracking systems.