Crack Detection Using Fully Convolutional Network in Wall-Climbing Robot

Abstract

Since the wall crack inspection of structures is difficult to access and to secure objectivity of the visual inspection, research on automatic inspection is being conducted. This paper is a study on the automatic detection of wall cracks of the wall-climbing robot, which aims to detect the cracks by robot itself in real time. Deep learning techniques are also applied to crack inspection, but there are difficulties in resource limitation in embedded environments. In this study, we examined the performance by experimenting with deep learning method that can be applied to embedded environment and the possibility of applying them to wall-climbing robot was presented.