How capture speed affected reconstruction

This note describes one project experiment with a video reconstruction that fragmented into disconnected components. Reducing the frame set improved that reconstruction. It does not establish a universal cause of fragmentation or an ideal frame count for other scenes.

Photogrammetry needs overlapping views with enough change in viewpoint to constrain geometry. In this dataset, densely sampled frames added many near-duplicates. Resampling to roughly 680 frames improved registration and processing time in the recorded comparison. Those results depend on this footage, its coverage, the camera motion and the processing configuration.

The experiment used a fixed frame budget. Under that particular workflow, a longer recording did not automatically contribute more images to reconstruction. That is a property of the sampling recipe, not a rule that longer footage cannot add detail. A different path, slower motion, sharper images or additional useful viewpoints can change the result.

For future captures, the project will check overlap, blur, changes in viewpoint and coverage before choosing a sampling rate. A fixed number of frames should not replace those checks. COLMAP’s image-acquisition guidance explains why viewpoint diversity, texture and overlap matter; it does not prescribe the project’s experimental frame budget.

Editorial clarification, 14 October 2026: limited the conclusions to the tested dataset and removed universal claims about normal video frame rates and recording duration.