Hyperspectral Band Selection Based on Fuzzy C-means and Dingo Optimization Algorithm

Dimensionality reduction is necessary for hyperspectral image processing to reduce redundant information and improve classification accuracy and efficiency. Band selection is a dimensionality reduction method that selects a representative subset of bands from a hyperspectral band set according to certain rules. A novel band selection method based on fuzzy c-means (FCM) and dingo optimization algorithm (DOA) is proposed, which uses the FCM Algorithm to divide hyperspectral image into several subspaces and then uses the DOA to select representative bands within each subspace. The objective function of the optimization algorithm integrates the band information entropy and the redundancy of the band subsets. Three commonly used hyperspectral image datasets were used to test the proposed algorithm, and support vector machine was used to classify the selected bands. The results have shown that the proposed algorithm has high classification accuracy compared with other algorithms, which verifies the effectiveness of band selection.