In hyperspectral remote sensing image band selection, there exist issues such as poor nonlinear separability, high redundancy, and the tendency of traditional optimization algorithms to get trapped in local optima. In an effort to tackle these obstacles, the research puts forward an improved band selection method based on the concept of Kernel Fuzzy C-Means Clustering Based on Adaptive Step Firefly Algorithm and Information Entropy Guidance (SSGIE-KFCM). The study achieves efficient band screening through a two-stage optimization framework, utilizing a Gaussian kernel function to enable high-dimensional mapping of band feature spaces and employing cross-sampling and information entropy-based grouping strategies for band feature extraction. Considering computational efficiency, an improved firefly algorithm (FA) is introduced to enhance the global optimization search performance of kernel fuzzy C-means clustering. Adjusting the step size of the FA effectively ensures its rapid convergence and the validity of individual position updates. The outcomes indicate that the proposed approach achieves an average band classification accuracy exceeding 90% on both the Indian Pines and Pavia University datasets, with an area under the curve value of 0.958, and consumes only 40% of the time required by traditional methods. Moreover, the improved algorithm proposed in the study exhibits superior discrimination performance across different ground feature bands, with spectral feature computation times of 0.058s and 0.172s, outperforming other comparative algorithms. The proposed method offers a lightweight solution for real-time processing of remote sensing hyperspectral remote sensing image and holds significant engineering value in agricultural monitoring and urban ground feature classification.
He et al. (Fri,) studied this question.