为了能够在视频监控、人机交互、视频会议等领域对人头部运动实施实时跟踪,给出了一种使用黑白摄像机对人平移或转身时的头部运动进行实时跟踪的方法.该方法主要由基于块特征的跟踪和基于头部几何特征的校正两个步骤组成 .块特征跟踪算法仅利用图象低层信息而不依赖于目标的具体模型,可实现对头部自由运动的跟踪.为解决块特征跟踪误差累积等原因造成的目标丢失问题,又采用了头部轮廓几何特征检验方法,根据跟踪窗口中头部轮廓位置的偏移来对块特征跟踪结果进行校正.另外,为提高转身运动时相邻两帧图象的特征跟踪正确率,还引入圆柱模型来拟合头部,并在展开柱面内进行块特征选取和跟踪.本文方法在P350微机上进行了实验,实验结果表明,系统能对长时间图象序列中人平移或转身时头部运动实施准确跟踪 .当跟踪窗口大小为120× 180pixels,块特征数目为80个时,系统的处理速度达到 30帧/s

This paper presents an improved real-time head tracking method, which can realize robust tracking of a person's translation and turning around with a black/white camera. This method consists of two main steps: block feature based head tracking, and head geometry based correction. The block feature based tracking only uses low-level image information and does not rely on the models of different objects, thus it can be used to track the free motion of the head. The correction step is introduced to address the shift caused by errors accumulating of block tracking. This step measures the displacements of the head contour in the tracking window of current image, and hereby to correct the results of block tracking. Moreover, to improve the performance of block tracking during person's turning around, we introduce a cylindrical model to approximate head, and extract and track features in the warped cylindrical surface. The resulting system provides robust and precise tracking over long sequence on a 350 megahertz microcomputer, and operates at 30 frames per second with the tracking window size of 120×180 pixels and the tracking sets of 80 features.