Research On Object Detecting And Tracking Methods For Visual Surveillance

The application of video surveillance system is increasingly being used nowadaysfor security, transportation, and military purposes. People’s safety consciousness isgradually developed in line with economic development. So far, the majority of theinstalled video surveillance system is still the traditional video surveillance systems, butthere are obvious defects in traditional video surveillance system-hardly being used todetect and track target invasion. So it can be used as ex-post analysis only throughreplaying the video. Intelligent monitoring system not only has the traditional videomonitoring system function, but also can detect and track target invasion in time, evenwith the ability in identifying target and analyzing the target behavior. Moving targetdetection and tracking technology is the basis for intelligent monitoring, and theresearch on this project shows important theoretical significance and practical value.The main difficulties in moving target detection and tracking technology lie in theneed to meet the quality requirements of real-time, robustness and accuracy, but thereare contradictions among these three aspects. So how to select a compromise plan mustbe taken into consideration. Scholars at home and abroad have published literatures totry to find practical moving target detection and tracking algorithms, which have certainlimitations, due to unpredictability of the environment, such as particular environmentin application, poor real-time of the algorithms, the problems in robustness and accuracy.Moving target detection and tracking technology were mainly studied in this paper, andimproved moving target detection and tracking technology can be gotten by analyzingand contrasting the original algorithms.The following work was completed in this thesis: a comparative study wasconducted on the image smooth filtering of video image, image enhancement, thresholdsegmentation, morphological processing and other image processing technology,whichwas used as the basis on detection and tracking of visual target; Moving target detectionmethod was put forward, based on the combination of Surendra background updatealgorithms and the difference background in the comparative study and summary of thetraditional moving target detection technology which combined with the technologyrelated to video image processing; A technical summary on a large number of movingtarget tracking algorithm was conducted, and a algorithm structure of moving targettracking was built basing on the integration method; The advantages and disadvantages of each were verified with the aid of experimental simulation, and the limitation of theapplication scope in tracking algorithm was pointed out, basing on contrasting theprinciple of SIFT, Kalman and Camshift tracking algorithm principle; Considering thatthe single feature matching tracking, SIFT Kalman filter and Camshift tracking canhardly meet the actual engineering on dynamic quality request, a kind of improvedKalman filter tracking and scale integration Camshift tracking algorithm in theintegration of moving target tracking algorithm was put forward, which was based onthe fusion technology, and finally, the desired effect was achieved under simulationexperiments.