Hi everyone,
I’ve got a lot of experience with photogrammetry and optimised image capture / image enhancement workflows, but I’ve been asked to record a very complex scene, and Reality Capture seems to be struggling. The images have enough overlap but the scene contains a lot of relatively plain walls and difficult lighting.
******* manages to align the images fine, but as usual it’s comparatively slow, so I’m wondering whether there are some settings I can tweak in RealityCapture in order to deal with these types of “difficult recording scenarios” (in this case mainly relatively texture-less areas). My normal RC align settings are attached (I believe these are the settings Milos generally recommends).
Which of these should I try:
- set image overlap to “low” so RC examines more of the outer edges of each image?
- increase the Max feature reprojection error to for instance 8?
- up the detector sensitivity to “high” or “ultra”?
- tweak the max features per image / preselector features settings?
- all of the above / other settings?
EDIT: for this particular scene the accuracy doesn’t necessarily have to be millimetric, as long as the scene looks good that’s fine (if that helps)
Thanks in advance for any input, I’m still in love with RC so it’s definitely not a critique, just trying to figure out how to further increase the likelihood of successful alignments.
Regards,
Thomas