Condition Monitoring on Railway Construction Site Using Timelapse Videos

• Published in: International Conference on Control, Automation and Systems (Online) pp. 1622 - 1627
• Main Author: Nicodeme, Claire
• Format: Conference Proceeding
• Language: English
• Published: ICROS 17.10.2023
• Subjects: Asset management; Condition monitoring; Image segmentation; Landslides; mask-rcnn; Object detection; Rail transportation; scene segmentation; Sensors
• ISSN: 2642-3901
• DOI: 10.23919/ICCAS59377.2023.10316899

Railway systems are usually in an open world environment. They are subject to climatic variations and their effects. A particularly disastrous event is landslides on tracks, which may cause heavy damages to the travelers and rolling stock. In response to those incidents, railway companies organize emergency work to clear the rails and to strengthen the earthen structure. During the reinforcement work, cracks may appear in the embankment, the landslide may spread, and boulders sometimes fall. To monitor and secure those areas, a timelapse device is installed. To date it provides hundreds of pictures a day, during daytime, as long as the construction work lasts to get feedbacks on the renovation process and train new agents. Data are also used to understand how events unfold through post-acquisition analysis by experts. The addition of Artificial Intelligence (AI) in the system could develop the possibilities offered by the system. In this paper, the authors present their AI-based tool for asset management and condition monitoring. Two lines are proposed in parallel. The first detects agents and machines while the second segments the environment for changes detection. Limitations and potential solutions are discussed, such as rain drops or fog blur, luminosity changes and images quality deterioration.