Research On Motion Detection And Its Application In The Intelligent Surveillance Field

Video surveillance systems have been widely used in many fields in recent years, but they are limited to record the scenes. An intelligent surveillance system that can automatically analyze the video and alert for possible anomalous behaviors would be greatly appreciated. An intelligent surveillance system aims to automate object detection, recognition, tracking and activity understanding in dynamic scenes that has the very great potential value in the enhancement of social security. As the first step of video processing in surveillance system, motion detection is one of the most important and difficult technologies.An intelligent surveillance system needs strong computing processors to deal with huge data in real-time. In digital surveillance system, DSP is a balanced choice on power consumption, flexibility, performance and cost. The series of ADI Blackfin are designed specially for lower power to meet the requirements of the applications of nowaday audio , video and communication. The processor combines DSP, RISC and video processing that is very suitable for research and development of motion detection based on video.This dissertation is dedicated to solve the programs of motion detection in intelligent surveillance system , including the design of hardware platforms, research and implement of the algorithms involved in motion detection.First is the design and implement of the video acquisition and processing system based on ADSP-BF533, including hardware selection, design of schematic and PCB map of the circuit,hardware debugging.Then aiming at the application in indoor scenes whose video is from a fixed camera, background which is relatively simple and changes slowly, this dissertation present a solution of motion detection based on background subtraction, including image preprocessing, simple background model initialization, motion detection, shadow detection and post-processing based on mathematical morphology. The major research focuses on the method of motion detection and shadow elimination. Successfully apply the YUV color space to L.Snidaro's background subtraction which based on HSV color space, use the brightness component Y of YUV color space and three-color to eliminate the shadow.Finally, through the massive work, the motion detection algorithm is transplant in the ADSP-BF533, and it has basically realized the mobile detection after optimization.Because of the large requirements of intelligent surveillance and other video applications, the research and implement of motion detection have very important significance both in theory and practice.