视频三维运动捕获系统中多运动目标跟踪方法  

Multi-targets Tracking Algorithm in a Video 3D Motion Capture System

作　　者：晏剑云[1] 郑江滨[1] 

机构地区：[1]西北工业大学计算机学院,陕西西安710072

出　　处：《微电子学与计算机》2008年第8期88-91,共4页Microelectronics & Computer

基　　金：国家"八六三"计划项目(2006AA01Z324);陕西省自然科学基金(2006F02);西北工业大学研究生创业种子基金(Z200752)

摘　　要：在视频运动捕获系统中,多标记点的匹配和跟踪是获取标记点三维运动参数的关键问题.文中提出了基于二维和三维卡尔曼滤波器相结合的三维匹配跟踪算法.该算法通过卡尔曼滤波器预测标记点在二维图像和三维空间中的位置,缩小了标记点的搜索范围,约束了标记点之间的匹配关系,在双目视觉下实现了标记点准确跟踪和匹配.实验结果验证了算法的有效性.Multiple markers matching and 3D tracking is one of the key problems in a video 3D motion capture system. In order to resolve this problem, we propose a 3D stereo matching and tracking algorithm based on 2D and 3D Kalman filter. These filters are designed to predict each maker＇s 2D coordination in the two video images and 3D space position, and these predicted position data implies the stereo vision matching restriction condition of each maker, so the correct matching and tracking benefit from the position restriction relationship of these makers. Several experiments demonstrate the proloosed algorithm＇s effecting.

关 键 词：人体运动捕获 卡尔曼滤波 三维匹配跟踪 多目标 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]