RailGuard: Vision-Based Railway Track Inspection and Fault Detection Using a Pole-Mounted Camera

Abstract

Railway tracks require frequent and reliable inspection because geometric irregularities, rail-surface defects, fastener failures, ballast deterioration, and objects on or near the track can affect safety, ride quality, and maintenance planning. Conventional manual patrols and specialized track-recording vehicles provide valuable measurements, but they can be labour-intensive, costly, and difficult to scale for frequent monitoring. This paper reviews contactless railway monitoring approaches based on computer vision, LiDAR, GNSS/RTK, inertial sensing, machine learning, ground-penetrating radar (GPR), and InSAR, and uses the findings to motivate an indigenous Integrated Track Monitoring System (ITMS). The proposed direction combines stereo or line-scan cameras, LiDAR, GNSS/RTK, IMU, an encoder, onboard edge computing, SSD storage, and real-time alerts. A practical system should not only detect a defect but also estimate its severity, associate it with a reliable track position, retain evidence, and communicate an actionable maintenance alert. The review highlights calibration, weather and lighting variation, vibration, limited labelled data, computing constraints, and the integration of detection with localization as key engineering challenges.