基于降维Gabor特征和决策融合的高光谱图像分类

针对传统高光谱图像分类算法忽略空间特征这个问题，提出一种基于Gabor特征和决策融合的高光谱图像分类算法。首先，通过系数相关矩阵智能地对相邻和高相关光谱带进行分组；接着，在PCA投影子空间中提取每组中的Gabor特征，以量化局部方向和尺度特征；然后，结合保留非负矩阵分解的局部性以减少这些特征子空间的维度；最后，对降维特征进行高斯混合模型分类，并使用对数分类池决策融合规则将分类结果合并。实验结果表明，所提算法优于传统和现有的共计八种先进的分类算法。

Hyperspectral image classification based on dimensionality reduction Gabor feature and decision fusion

Aiming at the problem of ignoring spatial features in traditional hyperspectral image classification algorithm, this paper proposed a hyperspectral image classification algorithm based on dimensionality reduction Gabor feature and decision fusion. Firstly, it intelligently grouped the adjacent and hypercorrelated spectral bands by coefficient correlation matrix. Then, it extracted Gabor features in each group from the PCA projection subspace to quantify the local direction and scale features. Then, it reduced the dimensionality of these feature subspaces by preserving the locality of the decomposition of nonnegative matrices. Finally, it classified the reduced dimension features by Gaussian mixture model, and merged the classification results by decision fusion rules. Experimental results show that the proposed algorithm is superior to eight kinds of traditional and existing advanced classification algorithms.