ATIC News
Researchers from the Advanced Transportation Infrastructure Center actively share their latest work at conferences. Faculty and students present papers on a variety of relevant topics, bringing UND's research to audiences across the transportation and civil engineering fields.
ATIC Heads to ACI Convention
Dr. Sattar Dorafshan will share UND's drone-based bridge inspection research with engineers from across the country in Atlanta this October. To view the schedule, visit the ACI Convention website. Select programming will also be available on-demand to attendees who participate remotely.
Title and Abstract
Quantitative Assessment of UAS Infrared Thermography for Bridge Deck Delamination Detection
Subsurface delamination in reinforced concrete bridge decks produces measurable surface temperature differentials that infrared thermography (IRT) can theoretically capture; however, translating raw UAS-acquired thermal imagery into reliable, spatially validated delamination maps remains an unsolved problem under field conditions. This study presents a quantitative, end-to-end UAS-IRT framework for bridge deck delamination detection, developed and validated against agency ground truth in service bridge decks in North Dakota. A UAS equipped with an IRT sensor was used to capture thermal images of bridge decks scheduled for repair. Computer vision techniques were applied to generate georeferenced mosaic thermal maps matching the deck’s geometry. A per-pixel source traceback system was embedded throughout the stitching pipeline, enabling any mosaic pixel to be traced to its originating image and coordinates.
The maps were co-registered with the NDDOT deck removal ground truth map, where delaminated portions of decks were repaired. Structured preprocessing, comprising ROI isolation, min-max normalization, Gaussian smoothing, unsharp masking, and CLAHE contrast enhancement, was implemented to reduce background-driven thermal variability by 55–60% while preserving structural thermal gradients. K-means segmentation (K=3) was applied to consistently isolated localized high-intensity anomaly zones with spatial extents of 5–25% of the deck ROI, exhibiting partial but inconsistent correspondence with ground truth removal boundaries. These results demonstrated that UAS-IRT delamination detection under field conditions is fundamentally constrained by acquisition timing, emissivity heterogeneity, and thermal gradient subtlety, and that georeferenced ground truth registration is essential for rigorous detection performance evaluation.