Video moving object tracking technology gradually becomes the study issue and gets wide applications in the military and civil fields etc. To extract video moving objects accurately and rapidly from the real scene, a novel method for auto-extraction is proposed. First, the adaptive threshold is used to get difference image between two adjacent filtered frames. And then, to reduce the influence of noise, the connected components of the binary image are labeled to determine the region of the moving object, and by comparing it with the spatial information by edge detection, the moving object model is initialized. Finally, the image is devided into several regions. The outer pixels of the model boundary are connected respectively and the close contours is finished. Then with the closed contours, the moving object is extracted. The experimental results show that the proposed algorithm can extract the single moving target with different speed and automatically as well as moving multi-targets.