Research And Implementation On Visual Surveillance System For Human Motion Target

Intelligent surveillance system for human motion target based on vision combines a lot of advanced subject technologies such as image processing, pattern recognition, artificial intelligence and computer control. The visual surveillance has been broadly applied in robot vision, visual pre-warning, intelligent traffic management, multimedia teaching and safety monitoring for important location. Meanwhile, detection, identification and tracking for moving target have the high academic value and research significance as one of hotest issues in recent years.Image pre-processing, detection, matching recognition and automatic tracking for human motion target based on visual image have been deeply analyzed. In the thesis, theoretical approach, simulation algorithm and experimental test have been integrated together and complemented each other. It laid a good foundation for the research and implementation on the intelligent surveillance system for human motion target. The main contributions of the thesis are:1. Scheme design and implementation on the hardware platform of system. The foreign and democratic development status and application key points have been summarized and analyzed. By comparing many visual surveillance systems, the hardware platform with PT (Pan-Tilt) and decoder was designed and realized in the thesis.2. Research and implementation on the image pre-processing and moving target detection. Firstly, the principle and implementation of image transformation, filtering and image enhancement based on histogram were studied.In the moving target detection, a new background update algorithm based on panorama was proposed. Theoretical analysis and experimental test have proved that this algorithm can realize the update of complex background with little computational complexity. Then image binary conversion, morphologic processing and edge detection for the target image were carried on to realize the moving target detection and segmentation.3. Research and implementation on the identification and tracking for human motion target. In the thesis, matching and identification based on human body feature were used by the selection and extraction of the feature and the identification principle establishment. The closed loop idea was adopted in the image detection and tracking system using centroid coordinate as the feedback quantity. A new PT control strategy based on constraint domain was proposed and fuzzy controller was designed. Automatic tracking was realized by the making of fuzzy control principle. At last, in order to implement fast real-time tracking, tracking based on Kalman prediction was used to estimate the target location of the next time. Both simulation analysis and experimental test confirmed the feasibility of the algorithm.4. Software implementation and practical application of the visual surveillance system. Based on the theoretical approach and experimental test analysis, video capture was realized by VFW (Video for Windows) and the BMP image data was loaded under Windows platform framework. Meanwhile serial communication was established by using MsComm control component. At last, "Garage Safety Surveillance System" and "Multimedia Video Teaching System" were designed and developed. Both the systems are on operation well in the laboratory with stable device and good generality.