Airborne Hyperspectral Imagery for Band Selection Using Moth-Flame Metaheuristic Optimization.

Area of Science:

• Remote Sensing
• Computer Science
• Artificial Intelligence

Background:

• Hyperspectral imagery contains numerous correlated spectral bands, increasing classifier training sample requirements.
• Selecting an optimal subset of bands is crucial for maintaining classification accuracy while reducing computational load.

Purpose of the Study:

• To investigate the efficacy of the Moth-Flame Optimization (MFO) algorithm for hyperspectral band selection.
• To evaluate MFO's performance in identifying optimal spectral bands for improved classification accuracy.

Main Methods:

• Moth-Flame Optimization (MFO), a metaheuristic algorithm inspired by moth navigation, was employed for band selection.
• MFO was tested on three benchmark hyperspectral datasets: Indian Pines, University of Pavia, and Salinas.

Main Results:

• MFO achieved high Overall Accuracy (OA) on the datasets: 88.98% (Indian Pines), 94.85% (University of Pavia), and 97.17% (Salinas).
• MFO demonstrated superior performance in OA and Kappa statistics compared to Particle Swarm Optimization, Grey Wolf, Cuckoo Search, and Genetic Algorithms.

Conclusions:

• The proposed Moth-Flame Optimization approach effectively addresses the hyperspectral band selection problem.
• MFO provides a robust and accurate method for selecting spectral bands, leading to high classification performance.