Wave frequency selection method for hyperspectral hyperspectral remote sensing image based on SSGIE-KFCM algorithm.

Area of Science:

• Remote Sensing
• Data Science
• Computer Vision

Background:

• Hyperspectral image band selection faces challenges like poor nonlinear separability, high data redundancy, and local optima in traditional algorithms.
• Existing methods struggle with efficient and accurate screening of relevant spectral bands.

Purpose of the Study:

• To propose an improved hyperspectral band selection method, SSGIE-KFCM, addressing limitations of traditional approaches.
• To enhance the global optimization search performance and computational efficiency in band selection.

Main Methods:

• A two-stage optimization framework utilizing a Gaussian kernel function for high-dimensional mapping.
• Cross-sampling and information entropy-based grouping for band feature extraction.
• An adaptive step firefly algorithm (FA) integrated with kernel fuzzy C-means clustering (KFCM) for improved global optimization.

Main Results:

• Achieved average band classification accuracy exceeding 90% on Indian Pines and Pavia University datasets.
• Demonstrated a 0.958 area under the curve (AUC) value, with processing time reduced to 40% of traditional methods.
• Exhibited superior discrimination performance and faster spectral feature computation compared to other algorithms.

Conclusions:

• The proposed SSGIE-KFCM method offers a lightweight and efficient solution for hyperspectral image band selection.
• The approach holds significant engineering value for real-time processing in agricultural monitoring and urban ground feature classification.