Evaluating the Precision of 3D Morphological Reconstruction for Asphalt Pavement Textures via Close-Range Photogrammetry

Abstract

Highlights • Dense-graded mixtures outperform open-graded ones in reconstruction fidelity.• Optimal photo counts are 17–25 for balancing accuracy and computational cost.• Texture indicators correlate with laser scanning but show systematic offsets.• Ssk and Sku are noise-sensitive and excessive photos may introduce instability. Abstract Accurate 3D reconstruction of pavement textures is vital for assessing surface performance. This study systematically evaluates the factors influencing the precision of close-range photogrammetry for pavement texture characterization. Four typical mixtures, AC-13, AC-16, SMA-16, and OGFC-16, were reconstructed from 17 to 145 photographs, with high-precision laser scanning as the reference. Reconstruction quality was quantified by point cloud density, elevation deviations, and four texture parameters. Findings reveal that dense-graded yields 81–87% of points within 0–0.5 mm deviation, while open-graded achieves only 51%. Point cloud density increases with photo count but follows diminishing returns, stabilizing between 21 and 73 images. Photogrammetric texture parameters systematically underestimate laser values, yet preserve relative ranking among gradations. Ssk and Sku are highly sensitive to extreme noise, especially on fine-textured surfaces. These insights offer practical guidelines for implementing photogrammetric techniques in pavement engineering.