Automatic UAV inspection of tunnel infrastructure in GPS-denied underground environment

In the Architecture, Engineering and Construction (AEC) industry, unmanned aerial vehicles (UAV) has been widely acknowledged as a promising tool to perform adaptive structural health monitoring automatically. However, there still remains some challenges for drones to collect image data of underground structures, primarily due to low light and no GPS conditions. In order to facilitate data acquisition, this article developed a mobile software development kit (MSDK) for drone using visual positioning and predefined controlling code, which enabled the drone to automatically fly along a designated sinusoidal route, whilst continuously taking videos and images of the tunnel surface. The developed MSDK was able to adjust the drone parameters (e.g., overlapping rate, inspection range, heading, flight direction between frames of the video) for different underground infrastructure conditions. Furthermore, a field test is conducted in an abandoned windless tunnel near Cork (Goggins Hill Tunnel) to test its feasibility. Results show that the 40-m difference between the designated routine and actual routine was 1.9%, and the collected data processed by Pix4Dmapper could reconstruct the complete tunnel scene and surface details. The navigation method proposed in this paper allows UAVs to perform automatic inspection without GPS, and the collected image data is used to build a tunnel panorama view