强背景噪声下红外目标的鲁棒性跟踪算法  

Robustness-tracking algorithm for the infrared target under complex background noise

作　　者：高国旺[1,2] 刘上乾[1] 秦翰林[1] 

机构地区：[1]西安电子科技大学技术物理学院,陕西西安710071 [2]西安石油大学光电传感测井与检测教育部重点实验室,陕西西安710065

出　　处：《西安电子科技大学学报》2010年第6期1098-1102,共5页Journal of Xidian University

基　　金：国家部委预研资助项目(41101050102)

摘　　要：提出了一种将边缘检测与改进Mean Shift算法相结合的红外目标跟踪算法.初选了原始红外图像边缘后,再利用非线性边缘检测算法进行处理,有效地消除了原始红外图像中的大部分噪声,并能获取高质量的图像边缘信息.在此基础上,采取更新目标模型、目标模板背景加权以及候选目标区域核加权的方式改进Mean Shift算法,以增强Mean Shift算法跟踪目标的稳定性及对背景噪声的鲁棒性,从而实现强背景噪声下运动红外目标的快速、准确跟踪.实验结果表明,该算法不仅计算量较少,提高了跟踪速度,而且对背景噪声有很强的鲁棒性.A tracking algorithm for the Infrared target is proposed that is the combination of edge detection and the improved Mean Shift method. After the edge of the original infrared image is detected roughly, the non-linear edge detection algorithm is presented that eliminates the most original image noise and could lead to a high-quality image. Based on this image, the improved Mean Shift algorithm that focuses on renewing the target model, background-weJgbtedness of the target template and Kernal Function-weightedness of the selected target region is applied to implement quick-tracking of a fastmoving target so that the algorithm is not sensitive to moving background noise, and thus it improves tracking procedure stability and robustness to background noise of the algorithm. Experimental results show that the combination of the nonqinear edge detection algorithm and Mean Shift tracking algorithm not only reduces the operand of algorithms and improves the tracking speed, but also has a strong robustness to background noise.

关 键 词：红外目标 边缘检测 Mean Shift方法 目标跟踪 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]