Loading...
Bridge Sub Structure Defect Inspection Assistance by using Deep Learning
Author(s)
Kruachottikul, Pravee
Cooharojananone, Nagul
Phanomchoeng, Gridsada
Chavarnakul, Thira
Kovitanggoon, Kittikul
Trakulwaranont, Donnaphat
Date Issued
October 1, 2019
Type
Conference Paper
Abstract
Road transportation is the most popular transportation in Thailand, which the top two highest traffic are the region-to-region highways; and then inter-city highways. Therefore, the regular maintenance is required to maintain the good condition due to road safety. The most significant process of bridge inspection procedures is sub structure inspection, which requires visual inspection as an initial step. This process is used to quick determine the damage severity i.e. appearance and crack that may cause damage to the structure strength. The current process requires that the experienced maintenance engineer to be on the field in order to visual inspect and estimate whether the maintenance is required. Yet, due to the limitation of number of expert engineers to be on the field, the photo verification is introduced to assist them so that they are no need on every inspection site. However, using human to verify has no standard and uncontrollable. They need to have experience and good knowledge. As well as it is highly depended on individual decision-making skill. Thus, in this paper, the deep learning technique will be presented to assist the expert for quality inspection process of bridge sub structure images. That is using image enhancement and then image splitting and overlapping for image pre-processing. After that applying CNNs for object classification. As a result, the total accuracy is 89% based on 3926 dataset.
Citation
2019 IEEE 10th International Conference on Awareness Science and Technology Icast 2019 Proceedings, 2019