Studies Motion Picture Detection And Human Gesture Recognition In Video Surveillance

Detection and recognition of moving objects in video surveillance systems is an important research topic in computer vision applications, multiple moving objects detection and identification is ongoing research hot spot. Video surveillance system for monitoring purposes is to monitor object in the scene of aberrant behavior detection and analysis, so it is very important significance for studying multiple moving objects detection and analysis of body posture. In this paper, study on multiple moving objects detection method and simple human gesture recognition method.Moving object detection algorithm is varied, different methods for different environments. Three commonly used methods is the background subtraction method, the frame difference method and optical flow method, these three methods have their own advantages and disadvantages. Background subtraction method can accurately detect moving objects, but it is more sensitive to environmental mutations or environment in the light of changing. Frame difference algorithm is simple and fast detection, but if the object’s movement speed is relatively large changes in the testing process, detection of object will empty or missed. Optical flow method less demanding on environment, light and the velocity of object, but this method of calculation is very large, it is used in a specific hardware system support. For the advantages and disadvantages of the above algorithm, this paper integration background subtraction method and frame difference method detects multiple moving objects, the effect of detection is stable, accurate and fast.Object recognition is the advanced processing of video surveillance systems, is the basis of behavior pattern analysis in video surveillance systems, and human gesture recognition is a relatively simple process of identification. This paper to stand, bend, to lay down three gesture recognition method research, there are three main steps in the recognition process:(1) establishment of the human body posture model;(2) people feature extraction;(3) human gesture recognition. In this paper, using Mean Shift algorithm to extract the human of contour image, through the object contour image to setting up the model of the human body posture. Using hierarchical identification to recognize human gestures after establish the human body posture model.