This project aims to inspect the brickwork and masonry assets of railway bridges, particularly the intrados of arches where access is limited. The project targets using drones to collect images autonomously under the arches and then analysing the images to automatically detect the defects in the structure. Among the defects to be detected crack, spalling, water seepage, insufficient mortar, misalignment, and crushing are targeted to be inspected through imaging. The intended system is envisaged to work in this way: one brings a drone to a starting edge of an arch and then the drone starts automatically scanning through the arch surface and collecting images. Later the collected images will be processed to detect the defects and inform the maintenance experts. The study will potentially make use of a mock-up bridge to develop the control and navigation of the drone and any available images of defects in order to develop image processing, and finally a verification will be run in an actual bridge arch.
1) Collection of images from the intrados of a bridge arc and identification of hazards. This work package develops a deep learning algorithm to detect a pre-identified set of hazards in the brickwork of the bridges. Previously collected images will be used for development of this intelligent image processing algorithm.
2) Control of a drone for autonomous navigation in front of a wall using proximity sensors and webcam. This work package relates to developing an autonomous controller for a drone, identification of suitable proximity sensors and a webcam-based image processing to detect and follow a wall, and integration this detection with the autonomous controller.
3) Autonomous navigation of a drone to scan the intrados of a bridge arc. This work package develops a wholistic system that detects the edges of an arc from inside, controls the drone from one edge to the other through a horizontal line, detects the other edge and makes the drone turn back following a parallel line on another horizontal line the same distance from the arch.