Intelligent video surveillance system must be able to detect video images of the target in real-time, which can then be the associated intelligence analyzed. Moving targets separated from the context which is change real-time quickly and accurately. And the further analysis and image processing can be done[4-5], the relevant attributes and features information and accurate identification of moving objects can be obtained, not only the basis decision analysis for the managers can be provided, but also intelligent management and control systems can be obtained. The motion objects detection algorithms include optical flow, inter-frame subtraction method and background subtraction method. In this paper, the main task is to achieve recognition within the pedestrian stops, considering the special nature of the scene inside the station and combining traffic analysis, along with the comparison of the characteristics and advantages and disadvantages of each algorithm, and decide what appropriate background model and updating algorithms to take, so the detection of pedestrians moving target can be achieved. Some issues such as the amount of calculating, interference and background updates can be addressed by the use of the characteristics of inter-frame difference and optical flow method