欢迎访问《兵工学报》官方网站,今天是 分享到:

兵工学报 ›› 2016, Vol. 37 ›› Issue (9): 1753-1760.doi: 10.3969/j.issn.1000-1093.2016.09.028

• 研究简报 • 上一篇    

基于鲁棒主成分分析的红外图像小目标检测

王忠美1,2,3, 杨晓梅2, 顾行发3   

  1. (1.电子科技大学 自动化工程学院, 四川 成都 610054; 2.中国科学院 地理与资源研究所, 北京 100101;3.中国科学院 遥感与数字地球研究所, 北京 100101)
  • 收稿日期:2015-06-04 修回日期:2015-06-04 上线日期:2016-11-04
  • 通讯作者: 王忠美 E-mail:ldwangzm2008@163.com
  • 作者简介:王忠美(1984—), 男, 博士研究生
  • 基金资助:
    国家自然科学基金创新群体项目(41671436); 国家“863”计划项目(2013AA122901)

Small Target Detection in a Single Infrared Image Based on RPCA

WANG Zhong-mei1,2,3, YANG Xiao-mei2 , GU Xing-fa3   

  1. (1.School of Automation Engineering, University of Electronic Science and Technology of China,2.Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China;3.Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100101, China)
  • Received:2015-06-04 Revised:2015-06-04 Online:2016-11-04
  • Contact: WANG Zhong-mei E-mail:ldwangzm2008@163.com

摘要: 鲁棒的小目标检测是红外目标搜索与跟踪的关键技术,提出一种改进的单帧红外图像小目标检测算法。该方法将原始红外图像通过预处理变换到新的红外块图像模式,在红外块图像上,将红外图像小目标检测问题转换为低秩矩阵和稀疏矩阵分离的鲁棒主成分分析(RPCA)问题。考虑到红外图像中噪声和杂波的存在,用交替方向方法求解带噪声的RPCA问题,获得稀疏目标图像,并对获得的稀疏目标图像采用简单的图像分割算法进行目标检测。对空天、海天、天云、海面4种不 同场景的红外图像小目标检测,进行仿真实验,结果验证了所提出算法的有效性。

关键词: 兵器科学与技术, 红外图像, 小目标检测, 块图像模型, 低秩矩阵恢复, 鲁棒主成分分析

Abstract: The robust infrared small target detection is one of the key techniques of infrared search and track systems. An improved algorithm is presented for small target detection in single-frame infrared image. The infrared image model is generalized to a new infrared patch-image model, and based on the new model, the small target detection is formulated as an optimal robust principal component analysis (RPCA) problem of separating low-rank and sparse matrices. Considering the presence of noises and clutter in infrared image, an alternating direction algorithm is used for solving the RPCA problem to obtain sparse target image, and a simple image segmentation method is used to segment the target image for target detection. The small infrared targets in different scenes, such as sky, sea-sky, cloud and sea surface, are detected and simulated. Experimental results demonstrate the effectiveness of the proposed method.

Key words: ordnance science and technology, infrared image, small target detection, patch image model, low rank matrix recovery, robust principal component analysis

中图分类号: