Advancements in Obstacle Intrusion Detection Methods for Rail Transit: A Comprehensive Review

Abstract

Developing intelligent obstacle intrusion detection systems (OIDSs) is crucial for the safety of train operations. Advances in graphics processing units (GPUs) and image sensing technologies have driven research and development of railway environmental perception using artificial intelligence (AI). However, compared with road transport, the current research and development efforts of railway OIDS are relatively insufficient. Although several papers have reviewed obstacle detection methods in railway applications, a comprehensive summary and discussion of the state-of-the-art AI-based methods is lacking. In view of this, this article aims to provide a thorough review of the traditional visual sensor-based and AI-based obstacle intrusion detection (OID) methods used in railways. This article first discusses the characteristics and advantages of various sensing devices, including cameras, light detection and ranging (LiDAR), and millimeter-wave radar, as well as the integration of these sensors. The focus is on rail area detection algorithms and obstacle detection algorithms, and the algorithm principles and model structures are summarized. In addition, this article provides an in-depth discussion of strategies for improving and optimizing the OID algorithms. Finally, the challenges and future requirements of OIDS in railway applications are outlined, and potential research directions are highlighted.