【作者】 刘兆英； 周付根； 白相志；

【Author】 LIU Zhao-ying,ZHOU Fu-gen,BAI Xiang-zhi (Image Processing Center,Beihang University,Beijing 100191,China)

【机构】 北京航空航天大学图像处理中心；

【摘要】 为解决多模图像的自动匹配并将之应用于红外目标识别,文中研究了基于小波不变矩和主成分分析(PCA)的多模态图像特征提取及匹配技术。首先,从红外图像特点出发对目标进行预处理,有效增强了目标,同时准确地分割出目标区域;其次,研究了基于小波不变矩的目标区域特征提取方法,并结合目标轮廓的形状特征通过PCA降低特征维数,实现红外目标有效特征的提取;最后,构造目标特征矩阵,实现特征匹配以及目标分类。实验结果表明,文中结合了目标小波不变矩特征和形状特征,并且采用PCA进行分析,实现了对目标特征的全面准确的描述,可以有效地用于提高多模图像匹配和目标识别的效率和精度。更多还原

【Abstract】 To solve the multi-modality image matching and infrared image automatic target recognition problem,this paper studies the feature extraction methods and the feature matching techniques based on wavelet invariant moment and PCA in multi-modal image.Firstly,pre-process has been worked out in order to enhance the target area and accurately segment the target area;Secondly,this article focuses on the extraction of target features based on wavelet invariant moment and the principal component analysis(PCA);Finally,feature matching matrix is constructed by the PCA to implement the feature matching.The experiment results show that the method proposed in this paper could effectively extract the features of multi-modal images,and achieve the goal of multi-modal image feature matching and target classification.更多还原