엣지 컴퓨팅 환경에서 객체탐지 기법을 활용한 도로 균열 탐지 시스템 개발

Abstract

In large cities with heavy traffic and many roads to manage, it takes months to even years to detect road cracks and repair them. One of main reasons for this prolonged maintenance job can be attributed to manual and batch processing of road crack detection. Manual and visual inspections are repeated from the initial crack detection task to the onsite re-diagnosis which cause significant delay in the maintenance process. Currently, some local governments are attempting to automate and accelerate those manual road crack detection and maintenance systems using artificial intelligence technologies and edge computing devices. In the field environment where the network connection is unstable, road crack diagnosis should be done without relying on the central server. During this process, the difference between the road scanner image and onsite acquired image should be properly handled by the edge device. In addition to this, the computational capacity required for training and operating artificial neural networks for crack detection needs to be properly distributed among the cloud server and edge devices. In this paper, to handle those problems of road management system, a method for developing mobile Internet environment over edge device that detects road cracks and supports repair works is studied.