Automation of pavement evaluation through examples based on 3D and AI technologies

Abstract

Pavement evaluation remains a fundamental component of transportation infrastructure management; however, current practices are often fragmented, labor-intensive, and limited in spatial coverage. Traditional methods rely on discrete measurements and outcome-based metrics, such as the Pavement Condition Index (PCI) and friction testing, which do not fully capture the underlying mechanisms governing pavement performance and safety. As a result, agencies lack continuous, objective, and scalable approaches for integrated condition and safety assessment at the network level. This study investigates the use of high-resolution three-dimensional (3D) pavement surface data and automated processing workflows to support standardized pavement evaluation. Two complementary case studies were conducted in Montana. The first utilized a 0.5 mm-resolution 3D imaging system to perform an automated PCI survey across 247.9965 miles of roadway in Gallatin County, with distress identification and PCI computation conducted in accordance with ASTM D6433. The second employed a 0.1 mm-resolution 3D laser system to collect statewide pavement surface texture data for macrotexture analysis using Mean Profile Depth (MPD) consistent with ASTM E1845. Results demonstrate that automated PCI computation from 3D data produces condition trends consistent with expected pavement behavior. Deterioration rates varied by functional classification, with local roads exhibiting the greatest decline (-6.37 PCI points per year), followed by minor (-4.55) and major roads (-4.22). Analysis of PCI versus treatment age confirmed a general decrease in condition with time, although variability within each classification resulted in low coefficients of determination (R squared is almost equal to 0.05-0.12), indicating that additional factors influence deterioration. Texture analysis revealed a right-skewed network distribution of MPD, with statistically significant differences across surface types (F is almost equal to 170.95, p is much less than 0.001). High-friction surface treatments (HFST) exhibited elevated MPD values and increased variability, while lower-texture surfaces were more uniform. The relationship between mean texture and variability provided additional insight into surface consistency and performance. Overall, this research demonstrates that high-resolution 3D sensing and automated processing can be integrated with ASTM-compliant evaluation frameworks to enable continuous, network-level assessment of pavement condition and safety. By linking distress-based condition metrics with texture-based surface characterization, this work supports a transition from outcome-based evaluation toward a more mechanistic understanding of pavement performance.