A Multi-Strategy Enhanced Bionic-Inspired Secretary Bird Optimization Algorithm for Numerical Optimization and...

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Image Processing

Background:

• The original Secretary Bird Optimization Algorithm (SBOA) suffers from limitations including insufficient population diversity, weak local exploitation, and susceptibility to local optima.
• Complex optimization problems and multilevel image segmentation require robust and efficient algorithms.

Purpose of the Study:

• To propose a Multi-Strategy Improved Secretary Bird Optimization Algorithm (MISBOA) that addresses the limitations of the original SBOA.
• To enhance the algorithm's population diversity, local exploitation ability, and convergence accuracy.
• To evaluate MISBOA's performance on benchmark functions and for image segmentation tasks.

Main Methods:

• Implemented a chaotic elite initialization strategy for improved initial population quality and diversity.
• Introduced an adaptive spiral Lévy flight strategy to balance global exploration and local exploitation.
• Incorporated a dynamic neighborhood-guided mutation strategy to maintain diversity and improve late-stage convergence accuracy.

Main Results:

• MISBOA demonstrated superior convergence speed, optimization accuracy, and robustness on IEEE CEC2014, CEC2017, and CEC2020 benchmark suites compared to other metaheuristic algorithms.
• Applied to Otsu-based multilevel threshold image segmentation, MISBOA achieved more accurate and stable segmentation outcomes.
• Performance was validated using metrics such as PSNR, FSIM, and SSIM.

Conclusions:

• MISBOA effectively overcomes the limitations of the original SBOA, offering improved performance in complex numerical optimization.
• The proposed algorithm shows significant potential for accurate and stable multilevel threshold image segmentation.
• MISBOA presents a promising approach for solving challenging optimization and image processing problems.