The Research On Motion Detection Technology Based On Flow Visualization

In recent years, since the increasing of public safety problem, the innovating ofhigh-performance micro-processors and the continuous developing of video analysistechnology, video surveillance system has been widely used in many industries, such assecurity, transportation and medical treatment. And intelligent visual surveillance haveachieved many functions。The key technologies of intelligent visual surveillance includeobject detection, object matching, object tracking, object recognition and video contentunderstanding, etc.Motion detection is the key technology of the intelligent visual Surveillance, it canprovide dependable dada for the high-level motion analysis and behavior understanding.This paper focuses on the technology of moving target detection in intelligent visualsurveillance system about how to achieve and extract the accurate position of the movingtarget effectively, and flow visualization technique is proposed for motion detection. Thispaper also do some exploration on the target matching algorithm, using gene algorithm toimprove the speed of covariance matrix matching. The main research contents in thispaper are as follows:(1) Moving targets and stationary background are mapped to different texture imagesusing the method of Line Integral Convolution which is based on flow visualizationtechnology,.(2)Using entropy tech to map the moving target texture to a low value, and thebackground noise texture to a high value, and then the OTSU algorithm is applied tosegment the areas of motion texture. We achieve to extract the moving target completely.(3)After the moving targets have been detected, author uses various characteristics ofthe target such as color, gradient to construct a feature model of covariance matrix, andthen use genetic algorithms to make the searching quick enough to locate the movingtarget in the image sequence.The experimental results show that the method in the paper can detect the movingtargets effectively and achieve the goal of targe matching and positioning in imagesequences.