在线咨询
中国工业与应用数学学会会刊
主管：中华人民共和国教育部
主办：西安交通大学
ISSN 1005-3085  CN 61-1269/O1
工程数学学报

工程数学学报 ›› 2015, Vol. 32 ›› Issue (1): 1-10.doi: 10.3969/j.issn.1005-3085.2015.01.001

• •    下一篇

基于奇异值分解的红外弱小目标检测

田  超,  高陈强   

  1. 重庆邮电大学通信与信息工程学院，重庆 400065
  • 收稿日期:2013-10-08 接受日期:2014-04-03 出版日期:2015-02-15 发布日期:2015-04-15
  • 基金资助:
    国家自然科学基金 (61102131)；重庆市自然科学基金 (cstc2014jcyjA40048).

Infrared Dim and Small Target Detection Based on SVD

TIAN Chao,  GAO Chen-qiang   

  1. School of Communications and Information Engineering, Chongqing University of Posts and Telecommunications, Chongqing 400065
  • Received:2013-10-08 Accepted:2014-04-03 Online:2015-02-15 Published:2015-04-15
  • Supported by:
    The National Natural Science Foundation of China (61102131); the Natural Science Foundation of Chongqing (cstc2014jcyjA40048).

PDF (PC)

10

摘要： 鉴于传统的基于单幅图像奇异值分解红外弱小目标检测算法的不足，提出了一种新的基于图像序列奇异值分解的红外弱小目标检测算法．首先，利用图像序列构造图像矩阵并进行奇异值分解，得到对应的特征值与特征向量；其次，利用处理后的特征值和特征向量重构图像序列，得到新的特征图像序列；再次，在特征图像序列中选取合适的特征图像进行处理从而增强目标并抑制背景；然后，对新特征图像进行阈值分割，得到要检测的弱小目标；最后，对序列中的每幅图像分别进行帧间位置修正与帧内位置修正，以达到检测红外弱小目标的目的．实验结果表明该算法具有很好的鲁棒性与实时性．

关键词: 红外图像序列, 奇异值分解, 弱小目标, 跟踪前检测

Abstract:

Considering the shortage of conventional single-image-SVD-based infrared dim and small target detection method, a new sequential-image-SVD-based method is proposed in this paper. Firstly, an image matrix is constructed upon sequential images and then decomposed by the SVD algorithm, and consequently the corresponding singular values and singular vectors are achieved. Secondly, new sequential feature images are reconstructed by using processed singular values and singular vectors. Thirdly, the selected feature images from sequential feature images are further processed to enhance the small targets and suppress the clutter background. Fourthly, the small targets are obtained by segmenting feature images. Finally, the detection task of small target detection is completed by refining the inter-frame and intra-frame target locations. Experimental results show that the proposed algorithm is robust and with real-time efficiency.

Key words: infrared image sequences, singular value decomposition, small target, detect before track

中图分类号: