Research On Behavior Recognition Of Intelligent Video Surveillance System

As extensive application of the video surveillance systems, Smart analysis of surveillance systems is getting more and more important. Detection and analysis behavior is an important component of smart video surveillance systems, it directly affects functional of the whole system.Therefore, detection behavior has become hotspots of the current study. present, the behavior detection and analysis methods are many, also without uniform standards, application of Various fields is different. Therefore, researching on a strong general methods of behavior analysis are very important.The human behavior recognizes is analyze and identify body movement patterns, it is widespread hotspots of research computer visual field. human behavior recognizes has a wide application value, It applied to the intelligent security systems, It can efficiently complete the task to guarantee safety and can save a lot of resources, it adds more economic to a society.First, this thesis introduce basis of the theory and techniques of behavior recognizes, also introduce method of objective tracking and detection methods, behavior recognizes. Then this thesis presents a strong method of behavior recognizes, and used for SVM method to complete behavior recognizes. We has a more in-depth research on human behavior detection and recognized technical in a static background, mainly completes the following works.(1) proposed a method of behavior recognizes based on Moving Track.able correct to recognize the Walking and running by the movement speed,and able correct to recognize the wandering by displacement vector,also achieve a method based on recognizing exceptional of direction. (2)proposed a method of behavior recognizes based on SVM, the height as characteristic value, use SVM categorize motion, felling, or downing can be categorized.(3) for all above the algorithms, this thesis carried out experiments on researching. The results indicate the method of this thesis proposed are more accurate and fast on human behavior detection and recognize.