Experiments with LiDAR and photogrammetry

Repeated capture of the same kinds of subject under different conditions — coverage, overlap, lighting, access — and compared what survived reconstruction.

Almost everything is decided at capture. A thorough, unglamorous pass reconstructs better than any amount of downstream cleanup applied to a rushed one.

Reflective, featureless and very fine detail remain the reliable failure cases.

Where is the point of diminishing returns — when does more capture stop improving the reconstruction?