Detection Of Human Abnormal Behavior By Video

As a popular effective method of security and defense technology,Intelligent video analysis and monitoring technology get more and more attention. Intelligent video surveillance is usually used in pattern recognition, computer graphics and image graphics processing, etc.,and use video surveillance system in which adds intelligent video analysis module, then using computer’s powerful computing and data processing capabilities selecting the screened video frame or interference information to automatically finding abnormal object, and then analyze and extract useful information among the video images,quickly and accurately locate the abnormal scene, timely determine the abnormal condition among the surveillance video screen, using the most excellent form to send warning signs, and then carry out an effective early warning, real-time processing,getting evidence of intelligent video surveillance systems timely and effective.This article outlines several methods to determine the current regular use of moving targets, such as background subtraction, the time difference and optical flow. By comparing the various methods, the paper mainly uses background subtraction method to detect moving targets. Background subtraction method is applicable to stationary cameras, can be completed on the moving target segmentation, very susceptible to events not related to the light, the weather changes and external, etc. Background image will be accompanied by time and changes in the environment changes, the establishment if no background model, it will cause changes in the background to detect moving objects have a huge mistake, individual background point is likely to be incorrectly classified to the front attractions. Thus in the process of moving target detection among modeling picture background played a crucial role, and therefore there is a need to create a suitable background model to adapt to the changing environment. Then proposed law Mean Shift for target tracking algorithm improvements. Secondly, this paper describes the classification method based on fuzzy clustering. Cluster analysis is a tool for testing data, based on the similarity measure or perception data classification many disciplines, test data for each class of model is more similar to the aggregation, so that the relationship between the modes have been estimated. In this paper, a graphic image, for example, the design of a pattern recognition system based on fuzzy clustering. After the experiment, the threshold at select appropriate conditions, to achieve the correct clustering of different patterns, and then analyzed by clustering whether abnormal behavior.