System Design And Implement Of Intelligent Supervisory Control Based On Human Activity Analysis

In recent years, along with video frequency monitoring technology unceasingly to intellectualized and digitized development, the intelligent supervisory control technology has already caused more and more domestic and foreign merchants and scholar's value and has conducted a series of research. The Intelligent Security Supervisory Control System based on computer vision not only completes the safeguard tasks efficiently, but also saves a great deal of human labor and efforts and provides as well as great economic benefits to the society.Focusing on the application of Intelligent Security Monitoring System, this paper designs a relevant human motion analysis system. It not only achieves to detect and track moving people in video sequences from a fixed background, but also achieves to recognize some exceptional human activities based on human motion information extracted from above steps. Firstly this paper introduces the research status and tendency of human motion analysis in the world detailed, and then analyses its application in Security Supervisory Control System. In this foundation, the work in the thesis makes three main contributions in the field:On the research of the moving objects detection and extraction under a static background, we summarize several algorithms currently and suppose an improved one. Combining the Background Subtraction and Temporal difference, we can get more precise moving-human contour. And then, a fast and simple method, named quadric binaryzation, is proposed to realize moving object segmentation and extraction.In the respect of moving objects tracking, a tracking algorithm fit for this system is introduced in detail, based on the analysis of current tracking methods. That is using the proposed region fusion method to mark positions of single or multiple moving objects. Then, Kalman filter is used to timely track moving objects in the scene.Human continuous motion is a sequence grouped by human posture. So we adopt Template matching method for human activities recognition. This paper regards each sampled contour of moving object as a basic action, and builds up a standard human activity sequence template. And then, the weighted Hu moment is used as feature values, the motion detected is compared with the template, and simple human activity recognition is achieved.