Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : May 27, 2020 - May 30, 2020
Crack inspections are regularly performed on concrete structures, are an important step in assessing the need for rehabilitation. In this paper, we proposed Machine Learning based automated crack inspection and Mixed Reality based visualization method. In the proposed method, human inspector can adjust undetected or erroneous detection of the inspection result, which can be used as ML training data. As a crack visualization method, we used Mixed Reality Head-Mounted Display (MRHMD) to show the inspect results visually, which fine cracks can be easily recognized from far distance. The proposed method confirmed the improvement of the accuracy of crack inspection by presenting new learning data to the system. Furthermore, we confirmed that displaying cracks using MRHMD has proved effective in showing fine cracks that human is hard to see.
