基于可区分性字典学习模型的极化SAR图像分类

Discriminative Dictionary Learning for Polarimetric SAR Image Classification

极化SAR图像分类是一个高维非线性映射问题,稀疏表示(CS)对于解决此类问题具有很大潜力。字典学习在基于CS的分类中起到重要作用。本文提出了一种新的字典学习模型,用于增强字典的区分能力,使其更适合极化SAR图像分类。提出的模型根据字典中两类子字典在分类中的作用对其相应的表达系数施加不同的稀疏约束。为使共同子字典能够抓住所有类共享的特征,对其相应系数施加稀疏约束,为使类专属子字典能够抓住类内独享的局部和全局结构特征,对其相应系数同时施加稀疏和低秩约束。由于共同子字典表达所有类共享的特征,我们以测试样本在类专属子字典上的重建误差作为准则进行分类。本文在AIRSAR的Flevoland数据集上对此算法进行验证,实验结果验证了算法的有效性。

Polarimetric SAR image classification is a high-dimensional nonlinear mapping problem, sparse representation （CS） has shown great potential to tackle such problems. The dictionary plays an important role in image classification. In this paper, we propose a novel dictionary learning model to obtain a discriminative dictionary. The proposed model is more suitable for polarimetric SAR image classification. According to the role played by two kinds of sub-dictionaries of the discriminative dictionary in the classification, this model imposes the different sparsity regularizations on the coefficients corresponding to such two kinds of sub-dictionaries. Specifically, to make the common sub-dictionary capture the features shared by all clas- ses, we impose the sparsity regalarization on the coefficients corresponding of it; to make the class-specific sub-dictionary capture intra-class intrinsic local and globe structure features, we impose sparsity as well as low rank regularization on the co- efficients corresponding of it. Due to the common sub-dictionary represents features shared by all classes, we use the reconstruction error based on each class-specific sub-dictionary for classification. The experiments are implemented using AIRSAR Flevoland data set. The experimental results confirm that the proposed method is effectiveness and improves the accuracy.