Multiobjective Sine Cosine Algorithm for Remote Sensing Image Spatial-Spectral Clustering.

Area of Science:

• Remote Sensing
• Computer Vision
• Optimization Algorithms

Background:

• Remote sensing image clustering is challenging due to high dimensionality and complex spatial-spectral information.
• Existing methods struggle with initial value sensitivity and local optima.
• Single objective functions inadequately model diverse remote sensing data characteristics.

Purpose of the Study:

• To propose a multiobjective sine cosine algorithm for remote sensing image data spatial-spectral clustering (MOSCA_SSC).
• To address limitations in evolutionary multiobjective optimization for clustering, specifically the balance between global and local search.
• To improve the accuracy and robustness of remote sensing image classification without prior information.

Main Methods:

• The clustering task is framed as a multiobjective optimization problem.
• Objective functions include Xie-Beni (XB) index and Jeffries-Matusita (Jm) distance with a spatial information term (SI_Jm measure).
• The sine cosine algorithm (SCA) is adapted for multiobjective clustering, incorporating a knee-point-based selection for updating solutions.

Main Results:

• The MOSCA_SSC algorithm effectively balances global and local search abilities in the evolutionary process.
• Experiments on UCI and real remote sensing datasets demonstrate the proposed method's benefits.
• The novel integration of SCA into multiobjective clustering shows promising results for spatial-spectral clustering.

Conclusions:

• MOSCA_SSC offers an effective approach for remote sensing image clustering.
• The method overcomes limitations of traditional clustering and evolutionary optimization techniques.
• This work advances the field of spatial-spectral clustering in remote sensing applications.