采用激活区域场景分析的红外目标跟踪算法  

An Activated Region Analysis Based IR Target Tracking Algorithm

作　　者：樊振华[1] 姬红兵[1] 杨金龙[1] 

机构地区：[1]西安电子科技大学电子工程学院,西安710071

出　　处：《计算机辅助设计与图形学学报》2011年第10期1741-1748,共8页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(60871074)

摘　　要：针对红外图像中目标易受杂波干扰的问题,利用协方差算子提取表观特征,在粒子滤波框架下提出一种基于激活区域场景分析的跟踪算法.首先通过建立激活区域划分出有效范围,降低计算量;其次分析目标所处场景并制定更新策略,避免了干扰对模板的影响;最后引入空间信息与表观特征融合对候选目标进行筛选,实现了跟踪窗的自适应调整,提高了跟踪精度.此外,提出一种正规化误差评价指标,改进了对跟踪效果的评价.实验结果表明,该算法对强杂波干扰环境下红外目标的跟踪具有良好的有效性和稳健性.To eliminate the clutter interference in IR videos, a new activated region analysis based target tracking algorithm is proposed in this paper. The algorithm is in the framework of particle filter and uses a covariance operator to extract the apparent features. Firstly, an activated region is established to determine the effective region and reduce the computational cost. Secondly, updating strategies are made with scenario analysis to avoid the adverse effects of interference on the template. Finally, the spatial information is fused with the apparent features to improve the tracking accuracy with an adaptive tracking window. In addition, a new normalized error evaluation criterion is proposed with improvement on the evaluation. Experimental results show that the proposed algorithm is robust and effective under heavy clutter environments.

关 键 词：协方差算子 激活区域 场景分析 红外目标跟踪 正规化误差 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]