The distributed nature of rail infrastructure makes it difficult to detect railway drainage assets and monitor their condition.

This project will explore the potential for a fusion of train-mounted sensors (such as GPR, laser scanning, and infra-red/optical camera systems) to identify key features such as drainage assets, wet beds, structures, flooding, and vegetation within the railway corridor. Initially this will be focused around earthworks, where early detection of landslips/slides, rockfalls and other related hazards is essential for avoiding accidents and ensuring smooth operation of passenger and freight trains. 

Advanced machine learning techniques will be developed based on a “few-shot” object detector to recognise and identify key features and hazards along a rail route, integrating with existing land, rail and weather data.