[4K3-J-13] AI application: sea and rivers

[4K3-J-13-01] Support Method for Diagnosing Degradation of the River Revetment using SfM and Semantic Segmentation

〇Junichiro Fujii¹, Yuri Shimamoto¹, Masazumi Amakata¹, Masahiko Matsukawa¹, Kazuya Akiyama¹, Hiroshi Obara¹ (1. Yachiyo Engineering Co., Ltd.)

Keywords:Structure from Motion, Deep Learning, Semantic Segmentation

The management of river revetment is mainly based on qualitative management dependent on human resources, and improvement of work efficiency is required. For this purpose, detection and diagnosis of the degradation needs to be performed based on quantitative data. We propose a method for automatically extracting absolute position and size of the degradation from images using SfM (Structure from Motion) and semantic segmentation, and confirmed its technical feasibility by experiment. It was shown that quantitative information necessary for the degradation diagnosis can be automatically obtained.