两级2维局部判别嵌入特征提取的SAR图像识别方法

SAR image recognition method based on two-stage 2DLDE feature extraction

提出一种基于两级2维局部判别嵌入(2DLDE)特征提取方法,并将其应用于SAR图像目标识别.该方法以矩阵的形式处理单个样本数据,对SAR图像采用两级特征提取,先后对图像矩阵从行列两个方向进行投影变换,避免了LDE方法将图像数据转化为向量带来的维数灾难和小样本问题,同时增强了特征判别性.结合相应的图像预处理过程和分类方法,应用两级2维局部判别嵌入特征提取方法对MSTAR SAR图像数据进行实验,证明了该文方法的有效性和优越性.

In the paper,a novel feature extraction method based on two-stage two-dimensional locality discriminant embedding was proposed and introduced into SAR image target recognition.The samples of SAR images were processed as the form of matrix directly and the features were extracted by two-stage method,where the image matrix was projection transformed from two directions of row and column respectively.Thus,the curse of dimensionality and the small sample size problem were avoided,which always appeared in LDE method where the samples were transformed to vector.Meanwhile,the features could be more discriminatory.Combined with the corresponding image pretreatment process and classification method,two-stage 2DLDE feature extraction was applied in the experiment on MSTAR SAR image data,and the results verified the effectivness and superiority of this method.