Hyperspectral Image Classification Based on Local Gaussian Mixture Feature Extraction

In order to further improve the classification accuracy of hyperspectral images, a classification method based on local Gaussian mixture feature extraction (LGMFEC) is proposed. The LGMFEC method first constructs a local neighborhood set for each sample based on the spatial structure of the hyperspectral image, and then extracts Gaussian mixture features from the local neighborhood set to fully characterize the spatial -spectral information and the related change information between them, and finally the local Gaussian mixture features arc integrated into a support vector machine (SVM) classifier containing a Riemann kernel function to complete the classification task. The experimental results of three sets of general hyperspectral datasets show that the classification performance of the LGMFEC method is better than several advanced classification methods to a large extent, especially when there arc fewer training samples.