Hybrid moving object detection system based on key frame extraction

In this paper we construct a hybrid moving object detection system. In this system, we first use the frame difference method to extract key frames in a given video sequence, then use the optical flow method and the HSV background subtraction method to extract the moving objects, respectively. We propose two hybrid methods: Improved Optical Flow method and Improved HSV Background Subtraction method. Experimental results have shown that our proposed system has strong robustness and be effective to detect moving objects in various conditions. We also introduce a human-computer interaction tool for selecting the focus area for users. This system will largely benefit for real-world video surveillance applications.