The method is applicable to hyperspectral remote sensing image classification. Provided is a hyperspectral remote sensing image classification method based on three-dimensional Gabor feature selection, comprising the steps of: A, generating a three-dimensional Gabor filter according to set frequency and direction parameter values (S1); B, carrying out a convolution operation on a hyperspectral remote sensing image and the three-dimensional Gabor filter to obtain three-dimensional Gabor features (S2); C, selecting, from the three-dimensional Gabor features, several three-dimensional Gabor features with contributions thereof to various classifications satisfying a requirement (S3); and D, classifying the hyperspectral remote sensing image by means of a multi-task sparse classification method using the selected three-dimensional Gabor features (S4). The method is based on three-dimensional Gabor features, and the adopted three-dimensional Gabor features contain rich local change information, and the features have strong capability of expression; the three-dimensional Gabor features are selected by means of a Fisher discrimination criterion, thus making full use of high-level semantics hidden between features, removing redundant information and reducing the time complexity in classification; furthermore, sparse encoding is performed, and three-dimensional Gabor features and multiple tasks are combined, thus greatly improving the classification precision.
展开▼