사진측량

A feature-based pipeline finds and matches identifiable scene points across overlapping photographs. Structure-from-Motion estimates camera poses and sparse 3D structure, then bundle adjustment refines the reconstruction by minimizing reprojection errors. Multi-View Stereo uses camera information to build denser geometry that can be meshed and textured. Image-only reconstructions do not automatically establish real-world scale and coordinates; surveyed ground control or other suitable references supply those constraints. Capture should provide overlapping viewpoints and matchable texture. Moving objects, weak texture, and changing reflections can undermine correspondences. Accuracy must be measured for the actual capture and control setup rather than inferred from a software name or an unqualified sub-centimeter promise.