Moving Target Detection And Statistics

The moving object detection is the basis of video analysis and has a wide application. This paper mainly focuses on the technology of the moving objects detection in image sequences of static background, and the technology of pedestrian counting in video which shot by gradient camera.This paper summarizes amount of mature algorithms, and studies the technology of the digital video constitution and camera calibration, and develops a pedestrian counting system for the video which is shot by gradient camera. This system utilizes data stream model which separates data and algorithm. The pedestrain counting system has three parts: pedestrian detection, pedestrian division and pedestrian counting. In the system, second order difference operator (SODP) is used to detect and locate the pedestrian, and a method which chooses the threshold for the difference image is proposed in the paper. The image segmentation results show the accuracy of the method. The pedestrian is divided by combining the SODP with edge detection, and the preliminary contour of the pedestrian is gotten, then the active contour is used to close the preliminary contour. At last, the system uses a detection bar to count the pedestrians. A video of the Tian'an Men square subway is used to test the system, and the precision of the pedestrian counting is higher than 80%.