I’m trying to assess reality capture as a replacement for other photogrammetry software. I will be creating models for engineering and heritage, so I’m looking for an error-correction workflow that results in an object with scale (provided through scale-markers) and measurements that can be tested and certified for accuracy.
My processes with other software was a multi-step workflow to slowly remove sparse-cloud interest-points with (relatively) high error, then re-optimizing/ re-aligning the camera positions to further refine the model. In using the trial version, I only found that selecting a lower “reproduction error” in the mesh-cloud settings as a way to help reduce error. While that helped, the lowest reproduction error reported was 0.55, even if I choose a lower number in settings. (Previously I aimed for a 0.3 reproduction error, on top of addition error-correcting methods)
Also, the only accuracy reporting is per-pixel, rather than a measurement. Is that because the trial version does not seem to have the scale feature?
Are there any other tips for an error-reduction workflow?
Please sign in to leave a comment.