Single image wire corrosion level identification using explainable AI

Shibin Parameswaran; Sara Wheeland; Ivan Febus-Rivera

doi:10.1117/12.3027476

9 May 2024 Single image wire corrosion level identification using explainable AI

Shibin Parameswaran, Sara Wheeland, Ivan Febus-Rivera

Author Affiliations +

Proceedings Volume 12950, Nondestructive Characterization and Monitoring of Advanced Materials, Aerospace, Civil Infrastructure, and Transportation XVIII; 129500U (2024) https://doi.org/10.1117/12.3027476
Event: SPIE Smart Structures + Nondestructive Evaluation, 2024, Long Beach, California, United States

Conference Poster

Abstract

Guy wires stabilize outdoor structures typically up to hundreds of feet in height. Exposed to potentially harsh environmental conditions, these wires can experience internal and external corrosion. Up-close visual inspection is often unrealistic given the height. A typical way to characterize corrosion involves taking samples of the wire, which requires destruction and replacement of the current wire. Non-destructive methods are preferred for this reason. To this end, we explore the feasibility of using computer vision techniques on images captured from Unmanned Aerial Vehicles (UAV) to automatically detect corrosion or swelling due to internal corrosion on a wire or cable. We leverage a data-centric approach to train a classifier for identifying corrosion level from a single RGB image. We evaluate the model performance on a dataset of wire images displaying 4 different corrosion levels. Finally, in order to provide additional insights and explainability to a human operator, we use GradCam to analyze the model's decision-making process and identify parts of an image that contributed to the resulting corrosion level label.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Shibin Parameswaran, Sara Wheeland, and Ivan Febus-Rivera "Single image wire corrosion level identification using explainable AI", Proc. SPIE 12950, Nondestructive Characterization and Monitoring of Advanced Materials, Aerospace, Civil Infrastructure, and Transportation XVIII, 129500U (9 May 2024); https://doi.org/10.1117/12.3027476

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

;

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE