| Intelligent Vision Counting is gradually applied to many fields as the advantages of liberating human beings from the heavy, long, repetitive work. Intelligent visual people traffic supervision system which human motion analysis is the focus of is in front of computer vision research in recent years Human detection and tracking compose the human motion analysis. As image acquisition equipments, or the limited installation location, or in order to avoid shading and other reasons, usually human-head region for detection is obtained. Also, intelligent visual counting in a static image process is widely used.Closely around the subject of intelligent visual counting, the methods are studied in this paper. Firstly, the detection and tracking of the human body in video sequences from top view were studied. And secondly, the formula is proved and a new approach is proposed to count holes in a static image. Finally, the design for human traffic detection system in intelligent visual surveillance system is proposed in the vertical stationary monocular camera environment. And I complete system module design, implementation of specific functions, and give real-time environment test results.In the body detection, firstly the video image noise is got rid of by Gaussian filter. Secondly, inter-frame difference is operated to extract moving objects, and holes in target regions are removed by morphological filter. Finally, for obtaining the clear human body contour, it is necessary to process the previous result image by canny edge detection. As the circular contour feature of human head, this paper gives contour model of human head, and then use Hough transform to detect human head. Hough transform itself is complex and spends large amount of calculation time. A new method of human head detection by gray gradient is proposed in this paper. For the detected human heads kalman prediction and data association based on objective matching methods are used to track multiple moving objects. |
