The Research Of Human Abnormal Behavior Recognition Based On Video

At present there are a lot of research on target detection and tracking method in the image processing field,experts at home and abroad also have solved some problems in the practical application of the algorithm,but there are still some defects about the algorithm at the same time,including the research methods of apply to special weather and the change of light in the video,these problems have hindered the effective application of the intelligent video monitoring system technology.The successful application on detection development research of abnormal behavior of people is not much.The traditional video monitoring have higher requirement to the camera and require more person on duty,it need added manpower and financial resources.This paper designed an intelligent detection system according to the defects of traditional video surveillance,the system can realize the real-time detection and timely alarm about the illegal intrusion and other anomalies,it also designed image enhancement algorithm according to the characteristics of nighttime,so that the ordinary camera could realize the nighttime monitoring,the system largely reduced the application cost of video monitoring and improved the monitoring function.So,this paper studies detection and tracking method of abnormal behavior of people in allusion to residential video and provided a stable design system for practicability and expansibility.Detailed research work are as follows:1 This paper analyzes the current demand for intelligent systems through the understanding of the research status at home and abroad,we design the response system architecture according to the needs and provide the conceptual design of the video system;2 We put forward the method of preprocessing image enhancement in the face of the situation of the foggy day and nighttime in the complex video scene,this method can maintain image color and saturation and can also process brightness in different regions,this algorithm effectively improved visual effect of image;3 We bring in the parameters of slow factor when testing the moving target,we analyze the local and global illumination condition,this article can adaptively update parameters in Gauss model,this paper introduces the gravity theory in physics into image processing and fused two methods of the gravity and the symmetric difference,the experiment results show that this algorithm can improve the problem of empty test target and defective edge;4 This article uses least squares to predict the testing tracking targets,it can improve operating efficiency of the system;5 Using the two parameters of barycentric position of the outlook moving target detection box and length-width ratio change of the external rectangle to realize the recognition of human abnormal behavior;6 We build the experiment platform of system on the PC machine according to the design scheme in this paper,each module function completes the debugging on the platform in the system,lately we optimized the interface platform and improved the convenience of human-computer interaction.