Datagov Opendata Platform

The rail fastening system is vital for securing the track on which trains run, significantly influencing the smooth and safe operation of trains. Currently, the Taiwan Railways Administration relies entirely on visual inspection by personnel, leading to the inevitable occurrence of oversights. In addition to gathering knowledge from overseas research on track inspection, this study also establishes equipment for collecting images of rail fastenings, including image recording and lighting equipment. Furthermore, utilizing the collected data on rail fastenings, the study employs image tagging and adopts the YOLOv4 deep learning model for training, followed by validation of fastening loss recall rates from test data. The results of this study can provide maintenance units of the Ministry of Transportation or the Taiwan Railways Administration with effective management of track safety in daily track inspections, serving as a reference for subsequent track maintenance and reinforcement.