Research And Implementation On Moving Object Detection Algorithm In Video Surveillance

With the development of science technology and standard living, traditional videosurveillance system can not meet the needs in production and livelihood. Intelligentvideo surveillance system has been widely used because of its intelligence characteristic,excellent real-time performance, high detection probability. Moving object detection isthe key technology in intelligent surveillance system. It is also the basis and kernel ofsubsequent identification, tracking, behavior understanding and behavior description ofthe target. In the paper, i designed and implemented a validation system of movingobject detection of video surveillance on the basis the research and improvement ofcommonly used algorithms.In this paper, functional needs of moving targets detection system is analysed. Onthe basis of the research of an application requirement of video surveillance,implementation framework is presented. Based on the research, implementation,contrast and analysis of frequently-used algorithm of moving object detection now, thealgorithm with dual background and multi-parameter based on inter-frame differencemethod and background difference method is designed and implemented. Firstly themoving foreground region is judged beforehand by the method of the uniformity ofinter-frame difference method and background difference method. Then backgroundimage is updated with different parameters according to the position of the obtainedregion. Finally, accurate moving object region is acquired by background differencemethod with updated background. By the analysis shadow feature in different colorspace and comparative study of detection algorithm, the existence of shadowphenomenon of moving object detection region is detected and removed by analgorithm for segmentation of cast shadows based on HSV color space. In the end, avalidation system of moving object detection of video surveillance is realized by thecode and integration of image preprocessing, moving object detection and shadowdetection. The system function is accurate, meets the requirement of design bysimulation testing with data file of surveillance video.Simulation results show the method presented in this paper solve the problem thatthe converting of an object between in a state of rest and in motion results in the“ghosting” by background difference method, and the problem that inside cavity or edgedefects take place in inter-frame difference method. It not only improves the accuracyand efficiency of detection remarkably, but also enhances robustness of the system. The results of the study play a role in the facilitative effect. in identification, tracking,behavior understanding and description of the target in application of video surveillance.