Research And Application On Object Classification Of Intelligent Video Surveillance System

With growing awareness of security, video surveillance with its intuitive, convenient, content-rich features has been widely popular. More and more cameras are deployed in areas such as traffic junctions, airports, shopping malls, parking lots and commercial buildings and other public places. These cameras are used to observe the monitored scene of persons and vehicles activities such as object. In the field of public safety computer vision applications have broad prospects as the help of intelligent video surveillance. Intelligent monitoring system developed into a popular research direction.Two cameras intelligent video surveillance system, fixed video camera is mainly used for collecting scene, and the video frame to pass motion detection module for processing; tracking camera capture close-up images of suspicious object information and this information is used for further face recognition.Moving object identification, classification and tracking is image processing, analysis, an important area of application. In the intelligent monitoring system, needs to analyze the movement of the object region detection, feature extraction, classification and identification with the movement of individual tracking. The key technologies of intelligent monitoring system is the object detection and classification, object tracking, object matching, object recognition and understanding of video content, and so on.Our work mainly included four points:1. In the static background of the video image, measure the true length and width of the ground object size and distance, speed and other parameters. 2. With true size of the object and other parameters, based on fuzzy set classification for moving object classification.3. In fuzzy classification of moving objects, achieve continuous time on the identification and confirmation by the transfer matrix of the object category.4. Experimental study of the content in Open CV platform. And further improved methods, so the results can be achieved practical application. Experimental results show that, using the object size parameter extraction and object classification, can be more accurately and quickly to achieve the object classification.