Quantitative Optical Nondestructive Evaluation of Bitumen-Aggregate Stripping Using Standardized Fluorescence Imaging...

Interfacial coating loss in bitumen-coated aggregates is difficult to quantify because the exposed aggregate regions are spatially heterogeneous, visually subtle, and sensitive to illumination and background conditions. This study develops a fluorescence-based optical nondestructive evaluation (NDE) framework for quantitative assessment of bitumen-aggregate stripping under standardized laboratory imaging conditions. A dataset of 5760 fluorescence images was collected from 120 physical BAP specimens representing 20 bitumen-aggregate combinations, with six viewing directions recorded for each specimen under eight acquisition environments. A lightweight no-reference image quality assessment model (LAR-IQA), together with visual inspection of shadow suppression and segmentation robustness, was used to select a reference acquisition protocol. The black-background, UV plus natural-light, glass-enclosure configuration provided stable contrast while reducing shadow interference. Three candidate segmentation methods were benchmarked against manually annotated reference masks, and the HSVSC algorithm achieved the best overall performance, with the highest mean Dice coefficient of 0.73. Using the standardized sensing-processing workflow, the image-derived stripping ratio distinguished material-dependent coating-loss behavior and revealed significant sensitivity to acquisition parameters. Aggregate type dominated the measured response under the tested conditions, with the mean stripping level ranked as limestone (3.94%) < quartz fine sandstone (4.17%) < basalt (4.95%) < granite (15.56%). ANOVA based on the 20 combination-level mean stripping ratios derived from 120 specimens, together with grey relational analysis, further indicated that, within the tested material set, aggregate-related descriptors were more strongly associated with the measured stripping variation than the selected bitumen descriptors. The proposed framework provides a repeatable, non-contact optical NDE route for converting subjective visual stripping assessment into image-derived quantitative measurement.