Research Of The Moving Object Detection And Tracking

As the key technology of intelligent video monitoring, moving target detectionand tracking are applied wildly in transportation and military and have become thehotspots of image processing. Background subtraction with the key technology ofbackground modeling and background update is the major detection algorithm ofmoving target detection. Modeling approach of single pixel background is focus onthe clustering and modeling process of single pixel and attracted more concern withthe advantages of robustness and real-time. Based on template matching of movingtarget tracking, appearance changes like covering, shading and so on will greatlyimpact the tracking effect. Common algorithms at present is not to distinguish fromboth yet, mainly due to the target template updating, and this will cause inaccuratematching because of the updating of interference information and will result in thefailure of tracking. This study includes:1. Background modeling and moving object detection: through the analysis ofthe single Gaussian model, Gaussian mixture model and codebook algorithm,Gaussian model of every pixel was found costs longer time and codebook modelconsumes more memory space. On the bases of above knowledge, this studyproposed for a single pixel, in RGB space based on pixel value clustering methodfor background modeling, there are three classes per pixel, dynamically updated atthe same time, thus saving the modeling time and reducing the memory footprint.Further by use of integral projection method to confirm the target movementinformation. The anticipated experimental results showed a higher stability androbustness of scene changes of the algorithm.2. Moving target tracking:Based on template matching of moving target tracking, appearance changeslike covering, shading and so on will greatly impact the tracking effect, especiallyhow to update the template become a hot spot currently. Shade and appearancechanges in the pattern matching method to realize tracking algorithm has a greatinfluence, Because shade and appearance changes has impact on the target’s pixelinformation, appearance change need to update the template, and shadow need not,or background information will be updated to the templates to cause the failure oftracking, distinguishing between shade and appearance change has a very important role. In addition, how to complete target tracking while target shadowing was alsostudied. In this paper, out-shadowed area is used to complete the target tracking,and the results of experiments showed that the algorithm had the ability to adapt tothe appearance changes and better processing capability with its capabilities tojudge the change of the target.