Abstract
After recent large earthquakes, the fast and safety emergency investigation of road and railway bridge systems is realized as one of the most important issue for the disaster mitigation and resilience of urban area in Japan. In this paper, the feasibility of bridge routine and post earthquake emergency inspection using small aerial photography Unmanned Aerial Vehicle (UAV) is verified by a series of onsite flight and investigation tests for 9 bridges. The basic performance of UAV for structure inspection are evaluated by compare the inspection result with the real bridge inspection reports. By using Deep Learning, it is possible to automatically detect damage and speed up image analysis. In this study, as a verification of the usefulness of Deep Learning for damaged images, image classification based on Deep Learning was performed using some images collected by past visual bridge inspection reports.
