Font Size: a A A

The Research And Implementation On Human Motion Tracking Based On Multi-Cameras

Posted on:2012-10-24Degree:MasterType:Thesis
Country:ChinaCandidate:Y LiFull Text:PDF
GTID:2248330395985690Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
In recent years, Computer3D animation has been the focus in the research areaof computer graphics,and motion capture technology adds new energy to it.Butbecause of the highly complexity of human motion,That is still the focus anddifficulty of study in real-time human motion capture accurately.Motion capture technology obtained the human motion from video sequences,and mapping the motion information to the new characters, making the motion of thetarget object similar as the video original motion. The purpose of motion tracking isto get the information of the human motion from video sequences,the performance ofthe motion tracking algorithms directly affects the accuracy and stability of themotion capture system,so it is the key to improve the performance of the motioncapture system that using a good robustness,high accuracy, real-time motion trackingalgorithm.The multi-carmera human motion tracking was reseached and the main researchachievement were as follows:In oder to get rid of manual annotaion human model in first frame,this thesispresented a automatic annotation human model method,and this method extracted thehuman model in images by using background subtraction method based on modifiededge,and then according to the prior knowledge of human model, human model jointswere indentified, using the theory of3D reconstruction and projection to mark thehuman model of each camera, at last the annotation results were corrected by the mostneighboring method. The experiment results shown that this method avoid the verbosemanual annotation, and achived the automaticly annotating human model of eachcamera effectively.A motion tracking method based on Multi-camera was presented in the thesis forcalculating the posture of human model,and the idea is:using the improved particlefilter algorithm to forecast the posture of human model, and the results were correctedby using the reverse epipolar line Geometry. The experimental data shown that thismethod improved the accuracy of human motion tracking, and partly solved theocclusion problem.A3D human motion capture system based on multi-camera was designed andimplemented, the system can achieve3D reconsturct the human model form image sequence,and the human motion data was formated in BVH files.
Keywords/Search Tags:Motion Capture, Motion Tracking, Motion detection, Particle Filtering, Reverse Epipolar Line Geometry
PDF Full Text Request
Related items