Effects of ground-based camera deployment parameters on SfM-Based 3D reconstruction of deformation features on stepped slopes

Stepped slopes are common in mountainous regions and are characterized by pronounced benches and risers, which generate strong occlusions during ground-based image acquisition and limit the accuracy of Structure-from-Motion (SfM)-based 3D reconstruction of surface deformation features. Despite the increasing application of SfM in slope monitoring, practical guidelines for optimizing ground-based camera deployment on stepped terrains remain limited. To address this gap, this study investigates the effects of ground-based camera deployment parameters on SfM reconstruction accuracy through geometrically scaled physical modeling and field validation. A 1:10 physical model of a rainfall-induced stepped slope was constructed to reproduce representative surface cracking patterns, with rainfall applied solely to induce deformation rather than treated as an experimental variable. Camera height, layout type, inter-camera angular interval, and camera number were systematically varied under controlled single-elevation configurations. Reconstruction accuracy was evaluated using the mean relative error between manually measured and SfM-derived crack widths. The results show that deployment geometry strongly governs reconstruction performance. A camera height of approximately one-third of slope height provides a favorable balance between spatial resolution and coverage. Compared with conventional horizontal layouts, fan-shaped configurations significantly reduce occlusion effects and improve crack-width measurement accuracy. Increasing the number of cameras generally reduces reconstruction error, but the improvement becomes marginal beyond four to five cameras under single-elevation deployment conditions. Under the optimal configuration, the physical model achieved a minimum mean relative error of 4.3%, while the field application at a prototype stepped slope yielded a mean relative error of 12.2%, which is acceptable for routine engineering monitoring. These findings provide practical and cost-effective guidance for ground-based SfM monitoring of deformation on stepped slopes in mountainous terrain.