Technology Of Tracking For Moving Object In Complex Scene Research

With the development of computer technology and image processing technology,moving object tracking technology has become an hot research spot in the field of computervision filed, which have wildly used in the aspect of intelligent transportation, securitysurveillance, robotic vision, military navigation, medicine and military guidance. Due to thereality of the scene is often very complex environment, when the object fast-moving,occlusion, deformation, light change conditions with a similar background and real-timerequirements which can still achieve stable tracking, it is hot spot of academic research andalso an urgent need to solve the problems in the process of practical application.To research the tracking algorithm when the object is in the tracking process, manyscholars rarely involve in the early image preprocessing of the tracking stage, but thepreprocessing of the video image is also very important for the late object tracking. In thestage of image preprocessing, this paper, firstly, has handled the video image noise filteringand enhanced, subsequently, has designed an color image enhancement algorithms basedHSV color model, the algorithm can highlight the image feature information and can be easilydetect prospect and background of moving object. As the problems of the backgrounddifferential object detection algorithm, through the analysis and research, so that a moreaccurate detection and extraction of the moving object.This article focuses on the problems of complex scene moving object tracking algorithmresearch and improvement. Firstly, the article makes a in-depth analysis on the Mean Shifttracking algorithm.View of the algorithm exists the shortcoming of when moving object sizechange track frame fixed lead tracking error and even loss the track object, through adjust thebandwidth of moving target, proposed on a improved algorithm to treat object scale adaptivechanges based on Mean Shift, the algorithm in the process of object motion in real-timechange the size of the tracking object is adjusted by the track frame such that lock the movingobject, to reduce the tracking error.Secondly, in order to solve the occlusion of object motion in complex scenes, the movingobject prediction problem has been researched based the Kalman filter. As classic Mean Shiftalgorithm lack of template update mechanism in the process of tracking object when theobject is obscured, design an anti-obscure tracking algorithm by fusion Mean Shift andKalman filter, in this algorithm, using the changes of the Bhattacharyya coefficient values todetermine the time occlusion occurs and occlusion end. The improved algorithm, due to the addition of target template updates mechanism, so to be able to more accurately describe thetarget feature to improve the accuracy and efficiency of the algorithm.Finally, to further verify the reliability of the article’ tracking algorithm, the articlecombine Visual C++6.0development platform environment with OpenCV vision librarydesign the modules of object tracking software system, implemented object trackingalgorithm in software, and through experiments to test the feasibility of the system, alsofurther validation the algorithm design of this article.