论文部分内容阅读
提出了一种新的基于图像块距离的邻域选择方法,并将其应用于流形学习中,得到一类新的高光谱图像非线性降维算法。该类算法利用高光谱图像物理特性,结合图像的光谱信息和空间信息,在最大限度减小图像信息冗余的基础之上,很好地保持了原始数据集的特性。与其它高光谱图像的降维算法相比,改进的流形学习算法不仅考虑到高光谱图像本身的空间关系,而且利用图像块距离更好地保持了数据点之间的局部特性,从而有效地去除原始数据集光谱维和空间维的冗余信息。实际高光谱数据的实验结果表明,所提出的算法在应用于高光谱图像分类时,与其它方法相比具有更高的分类精度。
A new neighborhood selection method based on image block distance is proposed and applied to manifold learning to obtain a new class of non-linear dimension reduction algorithms for hyperspectral images. The algorithm uses the physical characteristics of hyperspectral image, combines the spectral information and spatial information of the image, and minimizes the redundancy of image information, and keeps the characteristics of the original data set well. Compared with the other dimensionality reduction algorithms of hyperspectral images, the improved manifold learning algorithm not only considers the spatial relationship of the hyperspectral image itself, but also keeps the local characteristics between the data points better by using the image block distance, Remove redundant information from the spectral data and space dimensions of the original dataset. Experimental results on actual hyperspectral data show that the proposed algorithm has higher classification accuracy than other methods when applied to hyperspectral image classification.