Data supporting the publication: A Study on the Reconstruction of 3D Asphalt Surface Replicas for Statistically Reliable Analysis

This dataset supports the study <em>“A Study on the Reconstruction of 3D Asphalt Surface Replicas for Statistically Reliable Analysis”</em> and contains experimental, digital, and derived data used to develop and validate a methodology for replicating pavement surface macro-texture. The dataset includes measurements from physical asphalt cores, corresponding physical replicas produced via casting, digital replicas generated through computer vision, and calculated Mean Texture Depth (MTD) values for all sample categories.The study follows a four-phase methodology. In Phase I, pavement core samples representing five asphalt mixture types (including porous asphalt, dense asphalt, and rejuvenated surfaces) were collected from a Dutch highway network. Samples were cleaned and prepared to ensure accurate replication. No human or animal subjects were involved in data collection. The work was conducted within the Knowledge-based Pavement Engineering (KPE) program, a collaboration between Delft University of Technology, TNO, and Rijkswaterstaat, and complies with institutional agreements and infrastructure data governance requirements.In Phase II, both physical and digital replicas were created. Physical replicas were produced using silicone rubber casting followed by epoxy molding. A release agent was applied to preserve surface integrity, and curing conditions (temperature and time) were controlled to ensure consistency. Digital replicas were generated using smartphone-based video acquisition (4K resolution, 60 fps) under controlled lighting conditions. Videos were processed using Structure-from-Motion (SfM) techniques to reconstruct 3D point clouds. Post-processing included Poisson surface reconstruction, radius outlier filtering, mesh refinement, and geometric scaling to ensure consistency across samples.In Phase III, macro-texture was quantified using Mean Texture Depth (MTD). For physical samples, MTD was measured using the standardized sand patch test (ASTM E965). For digital replicas, an equivalent Finite Element (FE)-based MTD calculation method was developed. This involved discretizing the surface voids into tetrahedral elements and computing volume using determinant-based formulations, enabling direct comparison between physical and digital measurements.In Phase IV, statistical validation and comparison were performed. The dataset includes repeated MTD measurements (n = 110 per category) across four groups: Physical Core, Physical Replica, Digital Core, and Digital Replica. Statistical analyses include descriptive statistics, box plots, Shapiro-Wilk tests for normality, Levene’s test for variance homogeneity, and distribution similarity tests (Kolmogorov-Smirnov, Anderson-Darling, and Cramér–von Mises). Additionally, two-sample t-tests and Cohen’s d effect size analyses were conducted to evaluate differences in mean MTD values. An automated random masking and feature-matching technique based on computer vision (OpenCV) was implemented to reduce human bias in visual validation of surface similarity.The dataset enables reproducibility by providing raw and processed MTD values, reconstruction parameters, and validation outputs necessary to replicate the methodology. The results demonstrate that both physical and digital replicas preserve macro-texture characteristics with high accuracy (typically within 7% error), supporting their use in controlled laboratory studies of pavement performance.Data collection did not require formal ethical approval, as no personal, sensitive, or biological data were involved. However, the dataset originates from collaborative research involving public infrastructure assets, and its use is subject to institutional agreements and data-sharing restrictions to ensure compliance with legal and contractual obligations.