DETECTION OF CORROSION COVER PERCENTAGES IN STRUCTURAL MEMBERS USING DEEP LEARNING AND COMPUTER VISION

Date

2024-8-8

Author

Uzuncan, Hüseyin Dağlar

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

76
views

76
downloads

Cite This

Visual inspection conducted by inspectors has a great role in today’s structural evaluations. However, evaluations may differ through inspectors. Thus, to standardize the assessment of corrosion, this research tried to create a tool for site engineers. Basically, deep learning models were trained with new datasets of 1198 and 536 images, for corrosion and structural member instance segmentations. The outcomes were post-processed to increase the performance of the system. Detected member and corrosion instances were matched and corrosion percentages of structural members were calculated. The results were evaluated in terms of detected corrosion, member mask accuracy and the calculated cover percentage. In the end, overall accuracy was obtained as 52.3 % where averages of mIoU for corrosion, member segmentation and calculated cover percentage accuracy is equal to 67.4%, 93.4% and 82.9% accordingly. This research shows the possible usage of AI- powered tools for site engineers in corrosion rate detection where it is offered to public use with a website called “corrosionrate.org”.

Subject Keywords

Corrosion Detection, Member Detection, Instance Segmentation, Corrosion Rate, Structural Health Monitoring

URI

https://hdl.handle.net/11511/110853

Collections

Graduate School of Natural and Applied Sciences, Thesis