Polar lights optimizer: Algorithm and applications in image segmentation and feature selection

Abstract

This study introduces Polar Lights Optimization (PLO), an algorithm based on the aurora phenomenon or polar lights. The aurora is a unique natural spectacle that occurs when energetic particles from the solar wind converge at the Earth's poles, influenced by the geomagnetic field and the Earth's atmosphere. By analyzing the motion of high-energy particles and delving into the underlying principles of physics, we propose a unique model for mimicking particle motion. This model integrates gyration motion and aurora oval walk, with the former facilitating local exploitation, while the latter enabling global exploration. By synergistically combining these two strategies, the proposed PLO achieves a balanced approach to local exploitation and global exploration. Additionally, a particle collision strategy is introduced to enhance the efficiency of escaping local optima. To evaluate the performance of PLO, a qualitative analysis experiment is designed to assess its ability to explore the problem space and search for solutions. PLO is compared against 9 classic algorithms and 8 high-performance algorithms using 30 benchmark functions from IEEE CEC2014. Furthermore, we compare and analyze PLO with the current state-of-the-art methods in the field, utilizing 12 benchmark functions from IEEE CEC2022. Subsequently, PLO is successfully applied to multi-threshold image segmentation and feature selection. Specifically, a PLO-based multi-threshold segmentation model and a binary PLO-based feature selection method are developed. The performance of PLO is also evaluated using 10 images from the Invasive Ductal Carcinoma (IDC) medical dataset, while the overall adaptability and accuracy of the feature selection model are tested using 8 medical datasets. These results affirm the emergence of PLO as an effective optimization tool ready for solving real-world problems, including those in the medical field. The source codes of PLO are available at https://aliasgharheidari.com/PLO.html and other websites.