From Roads to Rails: Bridging the Gap in Signal Recognition Through Computer Vision

In this paper, we present a comprehensive review of traffic signal recognition through Computer Vision (CV) and its potential applications for operations and decision making within the railway domain. This methodology is well-established in the automotive industry (i.e., self-driving cars) where several specific methods have been developed, but little is written on signal recognition for autonomous trains or as Advanced Driver Assistance Systems (ADAS) for railways. While onboard Automatic Train Protection (ATP) systems currently dominate signal analysis in railways, CV-based traffic sign detection and recognition systems have significant potential for the industry. The objective of this study is to examine the state-of-the-art CV techniques applied to road signals, the challenges, and possible solutions for accurate and efficient signal detection and classification; and how these techniques can be incorporated into the railway sector. We also conduct a short literature review of traffic sign detection and recognition in both roads and railway systems. This study discusses CV systems utilized in road signals in terms of their suitability for railway applications and explores the advantages and challenges of utilizing these existing technologies for signal recognition in railway environments.