Human Tracking And Location In Home Environment

Video surveillance systems have been around every corner of human life, people’s research in the video surveillance technology is increasingly mature. Target tracking as a basic problem has been widely studied, scholars have proposed a variety of video object tracking methods, but also encountered some problems, such as non-rigid target gesture change in the environment, such as lighting changes. Video surveillance system in home environment is also facing such problems, especially the complex context of environmental impact on target tracking. In this thesis, target tracking in complex background environments was studied, the main contents are as follows:(1) For the movement detection problem in fixed background, this thesis analyzed several simple background models. Focusing on the codebook model, this thesis mainly studied on the classic codebook model, improved it to quickly adapt to the environment and proposed multi-mode blocked weights codebook model to achieve a rapid modeling and updating, in addition it can adapt to changes in light and other noises.(2) For the problem that target in complex environment, this thesis analyzed the particle filter tracking algorithm, especially the algorithm based on feature histogram. Target color histogram characterized the global feature of the target while texture histogram depicted the details of the target. According to the reliability of the two representations, integrate two histograms and achieve the accurate tracking of targets.(3) This thesis proposed a framework for tracking, divided the tracking problem into the main tracker, auxiliary tracker, target standard template library three parts. The main tracker using feature points matching method to ensure the accuracy of target tracking; when the main tracker fails, auxiliary tracker using particle filter algorithm to quickly retrieve the target, a combination of these two trackers ensures the continuity of target tracking; the standard object template library collects reliable templates to make sure that the target tracking reliability. In order to prevent tracking drift, this thesis designed an updating tactic to realize effective tracking.(4) Based on target tracking, this thesis realized binocular vision target positioning. Firstly, calibrate the camera. Then, use the artificial markers to calibrate binocular camera and compute the relationship between the two cameras. Finally, match the targets based on epipolar line constrain and find the same point to position the target.