[4Xin1] Poster session 2

[4Xin1-69] Study on Practical Application of Image Recognition for Rockfill Dam Embankment Inspection

〇Junichiro Fujii¹, Yukino Tsuzuki¹, Satoki Hirashima¹, Minoru Aihara¹ (1.Yachiyo Engineering Co., Ltd.)

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

Rocks called riprap are placed on the surface of rockfill dam embankments, and their inspection has been conducted by dam engineers through field surveys. The authors proposed an automatic detection method of weathered ripraps using deep learning object detection, enabling labor-saving inspection work and uniformity of inspection evaluation. In this study, a comparison was made between a case in which the proposed method was applied to the three weathering levels equivalent to the field survey, and a case in which the method was applied only to strong weathering, which has more distinct weathering characteristics and is important for dam management. Based on the results, a discussion of the practical application of the proposed method in dam management is presented, as well as future issues to be addressed.