一种新的空谱联合高光谱图像分类方法

A New Spectral-spatial Joint Method of Hyperspectral Image Classification

为了充分利用高光谱遥感图像的空间信息和光谱信息,以提高分类精度,文章研究了分别采用二维和三维Gabor滤波对高光谱遥感图像进行特征提取,然后与高光谱遥感图像的光谱信息进行融合,并基于堆栈式稀疏自编码器的深度学习网络对融合图像进行分类的方法。研究结果表明,所提出的空谱联合分类器与传统的光谱信息分类器相比,分类性能得到了显著提高,且三维Gabor滤波的空谱联合分类器的分类性能优于二维Gabor滤波的空谱联合分类器,并具有较强的鲁棒性。

In order to make the full use of spatial and spectral information of hyperspectral remote sensing image to improve the classification accuracy,this paper firstly uses two-dimensional and three-dimensional Gabor filters to extract the features of hyperspectral remote sensing images,and then fuses them with the spectral information of hyperspectral remote sensing images.Finally,it classifies the fusion images based on the deep learning network of the stack sparse auto-encoder.The results show that compared with the traditional spectral information classifier,the performance of the proposed spectral-spatial joint classifier is significantly improved,and the spectral-spatial joint classifier with three-dimensional Gabor filter is better than that with two-dimensional Gabor filter,and has strong robustness.