一种基于特征融合的运动目标跟踪算法  

A Moving Object Tracking Algorithm Based on Feature Fusion

作　　者：刘素珍[1] 邓和林[1] 

机构地区：[1]中国科学院光电技术研究所

出　　处：《激光与红外》2008年第8期837-840,共4页Laser & Infrared

基　　金：国家863计划高新技术预研课题(No.2003AA823050)资助

摘　　要：联合目标的灰度特征空间和梯度特征空间,构造了由目标的灰度加权直方图和梯度加权直方图联合表示的目标模型,利用均值平移算法在当前帧中迭代搜索目标位置。实验表明,该方法对于目标和背景灰度相似以及目标被部分遮挡时的跟踪是有效和稳健的。A new tracking scheme that is based on feature fusion is proposed. In this paper, by combining the gray space and gradient space, our target model is represented in the form of a weighting histogram using its gray feature and gradient feature, and frame-to-frame tracking is performed using mean shift. Experiments turned out the algorithm works efficiently and robustly for the occluding object tracking and the tracking when the gray scale of target and background are similar.

关 键 词：目标跟踪 特征融合 均值平移 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]