Photogrammetry

Core Idea

Photogrammetry extracts three-dimensional measurements from two-dimensional photographs by exploiting parallax -- the apparent shift of objects when viewed from different positions. When the same scene is captured from two or more viewpoints with known geometry, corresponding points in the overlapping images can be matched and their 3D coordinates computed through triangulation. Traditional photogrammetry uses precisely calibrated aerial cameras with controlled flight geometry, while modern Structure-from-Motion (SfM) photogrammetry works with unordered photographs from consumer cameras or drones, automatically solving for camera positions and scene geometry simultaneously.

Explainer

Photogrammetry is one of the oldest remote sensing techniques -- aerial photographs have been used for mapping since World War I. The fundamental principle is stereoscopic measurement: viewing the same scene from two different positions provides depth perception through parallax, just as human binocular vision does.

Traditional aerial photogrammetry uses precisely calibrated metric cameras mounted on aircraft flying systematic parallel flight lines with 60% forward overlap and 30% side overlap. Stereo pairs of photographs are processed in stereoplotters (originally optical instruments, now digital software) that allow operators to view the terrain in 3D and trace contours, buildings, and features. This remains the standard production method for topographic mapping and orthophoto production at national mapping agencies.

Structure-from-Motion (SfM) photogrammetry has democratized 3D measurement. SfM algorithms automatically detect and match distinctive features (keypoints) across large sets of unordered photographs, then simultaneously solve for all camera positions and a sparse 3D point cloud through bundle adjustment -- a least-squares optimization that minimizes reprojection errors across all images and points. Multi-view stereo (MVS) algorithms then densify the sparse cloud to produce millions of 3D points. The result -- a dense point cloud, mesh, orthomosaic, and DEM -- is similar to LiDAR output but derived entirely from photographs.

The combination of consumer drones and SfM software has made centimeter-resolution 3D mapping accessible for applications that previously required expensive LiDAR or aerial photography campaigns: construction monitoring, precision agriculture, archaeological documentation, mine volume calculation, and disaster damage assessment. The trade-off is that photogrammetry maps only visible surfaces and requires good lighting and texture, while LiDAR works through vegetation and in darkness.