论文部分内容阅读
为了增强高光谱遥感图像的分类效果,提出基于谱聚类和稀疏表示的两级分类算法。利用谱聚类将待分类的像元及其邻域内所有的像元分成两类,利用联合稀疏表示模型确定按规则选取的其中一类的具体类别,并以该类别作为像元的类。该算法充分利用高光谱图像的光谱及空间信息,两级分类过程均考虑了噪声及区域边界对分类效果的影响。进一步利用空间信息对分类算法进行修正,即关联邻近像元的类别,平滑分类结果。数值实验表明,该算法的分类精度高、稳定性好、抗噪性强。
In order to enhance the classification effect of hyperspectral remote sensing images, a two-level classification algorithm based on spectral clustering and sparse representation is proposed. By using spectral clustering, the pixels to be classified and all the pixels in the neighborhood are divided into two categories. The joint sparse representation model is used to determine the specific category of one of the categories selected by rule, and the category is used as the category of pixels. The algorithm takes full advantage of the spectral and spatial information of hyperspectral images. The two-level classification process considers the influence of noise and regional boundaries on the classification. The spatial information is further used to correct the classification algorithm, that is to say, to classify the neighboring pixels and smooth the classification results. Numerical experiments show that the algorithm has high classification accuracy, good stability and strong anti-noise.