Automatic Judgement of Steel Bearings Degrade by Deep Learning
In recent years, improvement of accuracy of image recognition by deep learning is remarkable.In this research, we examined the possibility of automatic judgment of bridge members damage by image recognition using Deep Learning for the purpose of improving the efficiency of maintaining and managing bridges with engineers lacking due to the declining birthrate and aging population.The verification was carried out on steel bearings where the contents of the photographs are hard to cause a difference due to the difference in the form of the inspection report because steel bearing have smaller features and characteristic shapes than the other members constituting the bridge.The type of damage to be considered was the coating deterioration that occurs most frequently among the damage that the steel bearing has.In addition, We also confirmed the improvement of learning accuracy when we learned from damaged images extracted only damaged members.Furthermore, we attempted to generate artificial damage image that can be used as a learning image using GAN (Generative Adversarial Networks).
Deep Learning Steel Bearing Coating Deterioration Bridge Inspection Automatic Judgments
Hitoshi Tatsuta Kohei Nagai Takanori Nomura
Nippon Engineering Consultants Co.,LTD., Tokyo, Japan The University of Tokyo, Tokyo, Japan Nippon Systemware Co.,LTD., Tokyo, Japan
国际会议
福州
英文
1189-1193
2018-11-04（万方平台首次上网日期，不代表论文的发表时间）