A Study on Excavator Detection to prevent gas lines digging accident based on Faster R-CNN and Drone/AR

Main Article Content

Don-Hee LEE, Young-Il Min, Jong-Sung Kim, Jeong-Joon Kim


Recently, damage accidents and damages during urban gas pipe excavation work have been increasing. Based on the Faster R-CNN AI model, an intelligent object recognition technique, excavators are detected in real-time images transmitted from drones and the excavation site is combined with GIS and Augmented Reality (AR) to monitor the excavator location after overlaying it on the map in real time. For intelligent architecture, Client Part is a drone, iPad app, and Server Part is designed as a GIS/AR and AI analysis model. Verification of accuracy was carried out by self-verification and on-site test bed verification. It has increased its ability to implement research by reflecting Real World's environment where regulatory sandboxes are applied. It has been confirmed that it is not unreasonable to apply to the site with an accuracy of about 94% and that the low detection rate, especially due to the nature of the industrial site, suggests that the research is successful. Approximately 58% of the time required for vehicle circuit inspection was reduced. This paper is expected to help develop safety management as the first case in Korea and abroad that combines drones with AI object recognition technology and GIS/AR technology into the urban gas safety management sector.

Article Details