STUDY ON ENHANCEMENT OF SANDBAR MONITORING BY DEEP LEARNING －TOWARDS ENHANCEMENT OF MANEGEMENT IN HOJO-RIVER－

Abstract

 In the mouth of the Hojo-River floodway that flows through the central part of Tottori prefecture, blockages occur frequently. River managers excavate sandbars to prevent flooding of internal and external water due to blockages. At that time, they judge the state of sandbar by checking the CCTV camera image. However, there are problems such as lack of personnel required for continuous monitoring and variation in judgment results.

 In this study, we develop a model that enables quantification and automation of the judgment using deep learning of artificial intelligence with advanced image analysis ability. In addition, we confirmed the usefulness of the model through field demonstration experiments aimed at improving sandbar monitoring. The model developed in this study may be applicable to rivers where CCTV cameras are installed in the direction of the river mouth.