Particle Swarm Optimization in Biomedical Technologies: Innovations, Challenges, and Opportunities

Introduction

Swarm intelligence (SI) is a population-based optimization approach, which is a sub-domain of soft computing techniques (Ab Wahab et al., 2015; Hamami & Ismail, 2022; Peška et al., 2019). It typically imitates the social behaviors of various animals observed in nature to solve problems. In quadratic equations, SI plays a crucial role in optimizing these equations with maximum accuracy and efficiency. One specific form of SI is Particle Swarm Optimization (PSO), which mimics the social behavior of birds while foraging (Huang et al., 2012; Xue & Shen, 2020). This article utilizes the expanded Kaptur’s entropy criteria approach as the fitness function to optimize two threshold values using the basic PSO methodology.

In computer vision, edge detection is a critical task and remains an active research area due to its frequent use as a preliminary step in extracting information from images. Edge detection plays a pivotal role in object identification, image segmentation, and understanding the structure of an image. The edges in an image provide valuable information about the frontiers or boundaries of objects within a computer vision application (Garcia, 2020; Haji et al., 2012; Islam et al., 2016; Pramanik & Bag, 2018; Sharma et al., 2014). These edges are formed due to changes in object characteristics such as geometry, shape, size, light intensity, and reflection. In gray-level images, an edge is defined as a line that separates two areas with different grey levels, which is a discontinuity in image intensity or its first derivative. There are two main types of image intensity discontinuities: step and line discontinuities. Step discontinuities are caused by rapid changes in image intensity, while line discontinuities are abrupt changes in intensity that return to their initial level after a certain distance (Azarbad et al., 2010; Gao et al., 2010; Ye et al., 2010). However, due to factors like smoothing from sensing equipment, low-frequency components in images, and external noise, these discontinuities are often not visible in real photographs, leading to the formation of ramp edges from step intensity changes and roof edges from line intensity changes.

Efficient image processing techniques, including clustering, are essential for accurate pattern recognition (Cheng et al., 2001; Deeparani & Sudhakar, 2021; Mishra et al., 2023). However, many existing clustering methods depend on human intervention and may not yield correct results when visual input is distorted. One of the most significant challenges in automation across various social sectors is the application of machine vision. This research specifically explores the use of machine vision in autonomous vehicles, focusing on identifying other road users (including autonomous vehicles), road infrastructure, and road markings (Liao et al., 2011; Nelson Jayakumar & Venkatesh, 2014; Wu et al., 2012).