ADVANCES IN MACHINE LEARNING FOR IMAGE ANALYSIS IN MECHATRONIC SYSTEMS: A CROSS-DOMAIN REVIEW AND RESEARCH ROADMAP

Abstract

Image analysis has become a fundamental component of modern mechatronic systems. It enables intelligent perception, decision-making, and autonomous operation across a wide range of industrial applications. However, existing literature remains largely fragmented, with most reviews concentrating on specific application domains or individual algorithmic developments, thereby limiting a comprehensive understanding of how machine learning-based image analysis is advancing across the broader mechatronics landscape. This review provides a cross-domain synthesis of machine learning techniques employed in image analysis for mechatronic systems. A comparative assessment is presented covering conventional machine learning approaches and deep learning methods. The study highlights the strengths, limitations, data requirements, and computational characteristics of these techniques. Major challenges affecting practical deployment are examined. Emerging research directions are also discussed, with particular emphasis on synthetic data generation, foundation models, self-supervised learning, multimodal intelligence, edge artificial intelligence, digital twin technologies, and trustworthy artificial intelligence frameworks. By consolidating current developments and identifying future research priorities, this review provides a structured reference for researchers and practitioners seeking to advance the development of intelligent, adaptive, and reliable mechatronic systems.