DAMAGE DETECTION OF CONCRETE SURFACE USINGANOMALY DETECTION METHOD BY DEEP LEARNING

Abstract

In recent years, technologies using big data such as deep learning have attracted attention for improving the efficiency of structure inspection. However, deep learning has a high degree of reliance on the quality and quantity of data, and the preparation of a data set on abnormal events used for supervised learning is a bottleneck in practical application. Therefore, in this study, for the purpose of complementing such problems, we examined and considered for the application to the detection of damage of concrete surface about the anomaly detection method by learning using the normal image. As a result of verification, it has been confirmed that the anomaly detection based on learning by normal images is effective for detecting the damaged area.