[4Yin2] Interactive session 2

[4Yin2-31] Application of Image Recognition AI to the Inspection of Rockfill Dams

〇Junichiro Fujii¹, Satoki Hirashima¹, Minoru Aihara¹ (1.Yachiyo Engineering Co., Ltd.)

Keywords:deep learning, image recognition, object detection, dam inspection

Riprap is placed on the surface of the embankment of a rockfill dam to protect the rock material inside. Since the riprap is weathered by repeated dry and wet weathering and freezing and thawing, it is necessary to understand the condition of the riprap, but the amount of work required for inspection by field survey is enormous. In addition, individual differences in judgments by dam engineers have been an issue. In this study, we applied object detection using deep learning to images acquired by autonomous UAV flight to determine the weathering status of riprap on the entire embankment in order to reduce the labor required for inspection and uniform evaluation of rockfill dams. As a result, it was confirmed that the system can comprehensively and uniformly evaluate the riprap of the entire embankment, although there is still a problem in unifying the criteria for creating training data.