Photogrammetry is the science of reconstructing three-dimensional geometry from multiple two-dimensional photographs of the same subject taken from different angles. Software analyzes feature points (corners, edges, distinctive texture patches) that appear in multiple frames, triangulates their positions in space using the differences in projection, and builds a sparse point cloud. From there, dense reconstruction, meshing, and texture projection produce a fully textured 3D model.
The technique predates digital photography by more than a century. Aerial photogrammetry was used for topographic mapping from the early 1900s, with two overlapping aerial frames analyzed in a stereoplotter to extract elevation contours. Modern digital photogrammetry uses the same principles at vastly higher fidelity. Leading software includes RealityCapture (now part of the Unreal Engine ecosystem), Agisoft Metashape, Pix4D, Reality Scan, and the open-source Meshroom. Each takes a set of photographs and outputs OBJ, FBX, or USD models that can be loaded into game engines, CAD systems, or VFX pipelines.
Capture technique matters enormously. The subject is photographed from every angle with substantial overlap (typically 60 to 80 percent) between adjacent frames so that the software has enough common features to solve relative camera positions. Diffuse, even lighting prevents harsh shadows from baking into textures. Consistent exposure across all frames avoids tonal mismatches in the final reconstruction. Reflective, transparent, or featureless surfaces (glass, polished metal, blank walls) defeat the algorithm because they offer no fixed feature points across frames; pros coat them in dulling spray or matting powder when the model permits.
Applications span industries. Surveying and civil engineering use drone-based photogrammetry to map sites with centimeter accuracy. Archaeology documents excavation sites in 3D so the dig record persists after artifacts are removed. Visual effects studios scan locations, props, and actors to integrate live action with CG seamlessly. Cultural heritage institutions preserve sculptures, buildings, and entire archaeological complexes against deterioration or destruction. Real estate, e-commerce, and game development now use photogrammetry to produce asset libraries of real-world objects rather than modeling them by hand. With smartphone implementations like Polycam and Apple’s Object Capture API, the workflow has reached consumers.
LiDAR-equipped phones and dedicated 3D scanners offer an alternative path, capturing depth directly via time-of-flight measurement. The two approaches are complementary: photogrammetry produces higher-quality textures from photographic data, while LiDAR offers faster and more reliable geometry, especially on featureless surfaces. Fused workflows that combine LiDAR geometry with photogrammetric texture are becoming standard in professional 3D capture.
Common pitfalls include shooting too few frames (alignment fails), changing the lighting mid-capture (textures mismatch), moving the subject between shots (the algorithm cannot resolve a non-rigid scene), and capturing a featureless area that the software cannot anchor. Raw capture and careful color management preserve the highest-quality input. The technique connects directly to newer developments like neural radiance fields (NeRFs) and Gaussian splatting, which use machine learning to produce novel-view renderings from the same kind of multi-angle photographic input.