基于双向2 DLDA与 LDA相结合的 SAR目标识别算法

摘要

线性判决分析（ LDA）用于图像特征提取时，存在着损失二维空间结构信息、计算复杂度大的缺点。二维线性判决分析（2DLDA）弥补了LDA的缺点，但2DLDA仅消除了图像各列间的相关性，所提取的图像特征维数仍然较大。为解决上述问题，采用双向2DLDA与LDA相结合的特征提取算法对图像的行和列同时进行压缩，减少特征矩阵维数，降低计算量。实验结果表明，所提出的SAR（ Synthetic Aperture Radar）图像目标识别方法有效地降低了图像数据维数，提高了识别率，并克服了方位角变化对识别结果的影响。%Linear discriminant analysis ( LDA) has the disadvantages of information losses in regard to two-dimensional spatial structure and high computational complexity when applying to image feature extraction.Two-dimensional linear discriminant analysis (2DLDA) makes up the flaws of LDA.However, it only eliminates the pertinence between the columns of image, while the number of features extracted is still large.In order to solve the problems, we adopt a new feature extraction algorithm to compress the columns and rows of image simultaneously, which combines the bidirectional 2DLDA with LDA, thus reduces the number of feature matrix dimensions and decreases the calculation amount.Experimental results show that the SAR images recognition method presented in this paper effectively reduces image dimensions numberand raises recognition rate, it also overcomes the impact on recognition result brought by the variation of target azimuth.