基于形态成分稀疏表示的红外小弱目标检测  

Infrared Dim Target Detection Based on Morphological Component Analysis Sparse Representation

作　　者：李正周[1] 王会改[1] 刘梅[1] 丁浩[1] 金钢[2,3] 

机构地区：[1]重庆大学通信工程学院,重庆400044 [2]中国空气动力研究与发展中心,四川绵阳621000 [3]中国科学院光电技术研究所,成都610209

出　　处：《弹箭与制导学报》2013年第4期29-32,36,共5页Journal of Projectiles,Rockets,Missiles and Guidance

基　　金：国家自然科学基金(61071191);重庆市科委自然科学基金(CSJC2011BB2048)资助

摘　　要：信号稀疏表示的超完备字典可有效感知信号的各种结构特征。针对红外小弱目标检测问题,文中提出了一种基于图像形态成分分析(morphological component analysis,MCA)理论的自适应信号稀疏表示的小弱目标检测方法。该方法根据红外图像信号自适应的训练和构造超完备字典,并进一步分为反映目标信号特征的目标子字典和表示背景噪声的背景子字典。然后求取待检测图像块在超完备字典的稀疏表示系数,挖掘目标和背景的稀疏表示系数差异,最后通过量化和比较信号在目标子字典的表示系数检测小弱目标。实验结果证明了该方法的有效性。The sparse representation of signals over redundant dictionaries can efficiently capture various characters or structures of signals. An efficient method based on morphological component analysis （MCA） was proposed for infrared dim target detection in this paper, com- bined with the self-adaption of the sparsity of signal. An adaptive dictionary was trained adaptively according to infrared image, and then the dictionary was subdivided into two categories ： the target dictionary which explains the target signal＇ s character and the background dictionary which embeds the background noise＇ s structure. Then sub-image blocks were extracted to seek its sparse coefficient over the adaptive dictionary. There is a significant difference between the coefficient of target and background noise. The target can be detected after a contrast of the sparse coefficients of the target dictionary between different blocks. The experiments show the approach is a practical and successful method.

关 键 词：小弱目标检测 稀疏表示 形态成分分析 自适应分类字典 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]