摘要
针对传统协方差矩阵跟踪方法中不能捕获目标旋转变化的问题,提出了一种基于椭圆协方差矩阵的红外目标辅助粒子滤波跟踪方法.首先对矩形协方差矩阵进行扩展,构建了椭圆协方差矩阵描述子,能有效适应目标的尺度和旋转变化,有效提高了目标模型的分辨能力.进而采用改进的李群结构来进行距离度量.在贝叶斯跟踪框架下,采用辅助粒子滤波采样粒子,解决了粒子滤波采样时由于没有利用观测值而造成粒子不能完全覆盖在目标位置附近的问题,最终实现了红外目标的准确定位.实验结果表明该算法简单有效,能准确跟踪尺度和旋转变化的红外目标.
To improve the capability of capturing object rotation for traditional region covariance matrix, improved infrared object tracking algorithm using elliptical region covariance matrix and auxiliary particle filter is proposed. Firstly, through extending the rectangle covariance region, the ellptical covariance matrix descriptor is constructed. Meanwile, the improved lie group structure is defined to build the similarity measure between the object model and candidate model. Finally, the accurate localization of infrared object is realized by the auxiliary particle filter. Experiment results verify the effectives and robustness of the proposed algorithm which can improve the tracking performance efficiently.
出处
《微电子学与计算机》
CSCD
北大核心
2013年第4期71-74,共4页
Microelectronics & Computer
基金
国家自然科学基金项目(61175120)
关键词
区域协方差矩阵
目标跟踪
李群结构
辅助粒子滤波
region covariance matrix
object tracking
lie group structure
auxiliary particle filter
