Research On Key Technologies Of Video Surveillance System

Video surveillance system (VSS) is one of new arising high-tech application fields. It spans many subjects including computer science, machine vision, image engineering, pattern analysis, artificial intelligence, etc. ASS can automatically analyse sequence image by the methods of computer vision and video analysis. When abnormity happens on the location which is monitored, the system automaticly alarms. Criminality can be avoided by using intelligent monitoring system with less manpower, money and material resources.However, there are still many problems in the fields of both theory research or in applications. Large numbers of researchers devoted themselves in the area and have already achieved many progresses. The dissertation studied the some key technologies of VSS based on the current conclusion. The main content of this dissertation can be summarized as follow ings:1) Threshold adopted is a key to the moving objects detection algorithm based on difference of frame. The algorithm which connects background subtraction and two consecutive frames subtraction to detection often goes wrong warning message, just uses an operation on experiential threshold. So the algorithm based check of self-adapt threshold is proposed, and overcomes shortcoming of traditional algorithm.2) An method to detect image edges based on multiscal wavelet transform is proposed in the thesis. A digital image is firstly decomposed with multiscale wavelet transform. Detailed section of every scales are detected with zero crossings by Marr edge detection operator. Finally, reconstructed image edges of every scales in terms of actual expectation. Experimental results have verified the veracity of solving contradiction of hold down noise and abstract edge with the method. Single-pixel detected with zero crossings by Marr edge detection operator can reduce complex of calculation in the next procedure of objects tracking.3) A moving objects tracking algorithm is proposed based on reformative Hausdorff distance matching . This method which uses Kalman filter's predictionto estimate direction of moving objects can reduce scope of character search and enhance speed of calculation. On the ground of accordant rule, A tactics based on connecting directive Hausdorff distance matching and average Hausdorff distance matching to tracks moving objects is proposed. Experimental results demonstrate that this method can not only provide more robust occlusion handling under the circumstances of objects disappeared or background shifted but also decrease the processing time in performing motion object tracking.