基于可区分性字典学习模型的极化SAR图像分类

发布时间：2018-06-19 17:46

  本文选题：极化SAR图像分类 + 超完备字典　； 参考：《信号处理》2017年11期

【摘要】：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。
[Abstract]:Polarimetric SAR image classification is a high dimensional nonlinear mapping problem. Sparse representation (CSS) has great potential to solve this problem. Dictionary learning plays an important role in CS-based classification. In this paper, a new dictionary learning model is proposed, which is used to enhance the distinguishing ability of the dictionary and make it more suitable for polarimetric SAR image classification. The proposed model imposes different sparse constraints on the corresponding expression coefficients according to the role of two sub-dictionaries in the classification of dictionaries. In order to make the common sub-dictionary grasp the characteristics shared by all classes, and to impose sparse constraints on the corresponding coefficients, the class specific sub-dictionary can capture the local and global structural features that are unique to the class. Both sparse and low rank constraints are applied to the corresponding coefficients. Because the common sub-dictionary represents the characteristics shared by all classes, we use the error of the test sample reconstruction on the class specific sub-dictionary as the criterion for classification. The algorithm is validated on the Flevoland dataset of AIRSAR. The experimental results show that the algorithm is effective.
【作者单位】： 武汉大学电子信息学院;
【基金】：国家自然科学基金项目(61771014)
【分类号】：TN957.52
，

本文编号：2040740