基于稀疏表示和词袋模型的高光谱图像分类

Classification of Hyperspectral Image Based on Sparse Representation and Bag of Words

为增强高光谱图像稀疏表示分类方法中稀疏字典的表征能力并充分利用高光谱图像的光谱信息和空间信息,提出了一种新的基于稀疏表示和词袋模型的高光谱遥感图像分类方法。首先利用词袋模型算法结合高光谱遥感图像数据集生成各类别专业码本,作为字典中对应的原子构造稀疏表示字典。在计算每个像元的对应稀疏表示字典中的稀疏表示特征时,利用空间连续性约束对像元的稀疏表示系数进行空间维的约束。最后根据最小重构误差实现高光谱图像分类。高光谱遥感数据实验结果表明:所提方法能有效提高分类效果,并且其分类精度和Kappa系数都优于其他稀疏表示方法以及单独使用光谱信息的方法。

To enhance representation ability of the sparse dictionary for hyperspectral image classification using sparse representation and make full use of spectral information and spatial information of hyperspectral image, a novel hyperspectral image classification method based on sparse representation and bag of words was proposed. First, some professional dictionaries of each class are generated by bag of words algorithm based on the hyperspectral remote sensing image dataset, and the sparse representation dictionary is obtained by merging these professional dictionaries. Then, the sparse coefficient of each pixel is calculated according to the sparse representation dictionary, and the spatial continuity is used to constraint the coefficient by using the information of its neighborhoods. Finally, the classification of the objects is determined by computing the minimum reconstruction error of it on each professional dictionary. Experiments on hyperspectral remote sensing images indicate that the proposed method has better performance, a higher overall accuracy and Kappa coefficient than other sparse representation methods and the method based on spectral information respectively.