改进的联合稀疏表示算法应用于高光谱图像分类

Improved Joint Sparse Representation Algorithm Applied to Hyperspectral Image Classification

摘要: 高光谱遥感图像含有大量光谱和空间信息，但同时存在数据冗余和噪声干扰问题。为了解决上述问题，本文提出一种高光谱图像分类方法：改进的联合稀疏表示算法(A2-JSRC)。该算法通过计算挑选与中心像元相似度高的邻域像元来构建联合稀疏表示模型，利用SOMP算法提取出稀疏特征并获得初步分类结果。随后，采用邻域内投票的方式减少错分点，将数量最多的类别确定为中心像元的最终所属类别。通过在不同数据集上进行实验对比，可以看出本文提出的算法优于一些传统算法，能够极大地提高分类精度，使实验结果图与真实地面图更加接近。

Abstract: Hyperspectral remote sensing images contain a lot of spectral and spatial information, but there are also data redundancy and noise interference problems. In order to solve the above problems, this paper proposes a hyperspectral image classification method: an improved joint sparse representation algorithm (A2-JSRC). The algorithm constructs a joint sparse representation model by calculating and selecting neighboring pixels with high similarity to the center pixel, using the SOMP algorithm to extract sparse features and obtaining preliminary classification results. Subsequently, the method of voting in the neighborhood is used to reduce the wrong points, and the category with the largest number is determined as the final category of the center pixel. Through experimental comparison on different data sets, it can be seen that the algorithm proposed in this paper is superior to some traditional algorithms, can greatly improve the classification accuracy, and make the experimental result map closer to the real ground map.