Optimal Camera Angle for Measuring Joint Roughness Using UAV-Based 3D Photogrammetry

Abstract

ABSTRACT: Joint roughness strongly affects rock mass engineering properties, including shear strength and hydraulic behavior. Integrating 3D photogrammetry with unmanned aerial vehicle (UAV) technology rapidly generates 3D roughness profiles from 2D images. However, inaccuracies in roughness measurements can occur due to survey errors, and the directional nature of roughness complicates the analysis of how these errors affect the inaccuracies. This study simulates joint roughness coefficient (JRC) measurement distortion due to UAV-based 3D photogrammetry errors, considering image distortion, joint plane orientation, and JRC direction. We measured JRC values through 3D photogrammetry on joint planes of a rock mass at Mt. Gwanak, Seoul, and compared them with JRC values distorted through the simulation. The comparison result shows that the simulation has a high performance in predicting JRC errors with an RMSE of 1.23. Based on the simulation, we propose guidelines for camera angle and UAV movement direction to minimize JRC measurement errors. The study recommends that flying the UAV along the strike of a joint plane with a camera angle orthogonal to the plane can minimize JRC measurement errors when the roughness of interest corresponds to the dip vector of the joint plane. 1. INTRODUCTION Joint roughness has a significant impact on both the mechanical and hydraulic behavior of rock masses (Paixão et al., 2022), and it should be a critical factor in designing rock structures. The Joint Roughness Coefficient (JRC), one of the most commonly used roughness parameters in practice (Barton, 2023), can be physically assessed through tilt tests (Barton and Choubey, 1977) or estimated visually from roughness profiles. Since measuring JRC values from visual inspection can be subjective, research has also been conducted using statistical parameters such as fractal dimension (Tse and Cruden, 1979), maximum amplitude of the profile (Barton and Bandis, 2017), and the standard deviation of profile gradients Z2 (Myers, 1962). The methods of obtaining roughness profiles can be categorized into two methods: contact and non-contact methods (Ge et al., 2014). Contact methods such as linear profiling and stylus profilometry are advantageous over non-contact methods in terms of cost-effectiveness, but they are time-consuming and unsuitable for surveying large areas (Paixão et al., 2022). When investigating large rock surfaces, non-contact methods like 3D photogrammetry or LiDAR (Light Detection and Ranging) technology are preferred for acquiring 3D roughness profiles due to their efficiency compared to the time-consuming contact methods.