Development of the automatic detection system for Sabo facilities using photos taken by UAV

Abstract

Inspection of Sabo facilities in mountainous areas has been done manually, but in recent years, some of the work has begun to be done by UAVs. Various methods have been proposed to detect anomalies in Sabo facilities from videos and photos taken by UAVs. However, to reduce costs and shorten detection times, it is necessary to improve AI-based automatic detection technology. In this study, we used UAVs to take photos of Sabo dams at different altitudes across various seasons, focusing on the Sabo dams of the Nikko Sabo Office. Also, we attempted to detect abnormalities by taking the difference between photos taken at two different times. This study has shown that the accuracy of detecting anomalies has been improved by using images from two periods of the same season, using an oblique angle of view, and taking images from a long distance.