Combine photogrammetry and laser data for better registration

[DEVR2098]

I don’t know what goes on”under the hood’ of recap when it comes to associating scan data and the photographs that the blk360 produces, so flame me if it’s already doing this but it seems that there are two competing approaches to reality capture. Approach one- -laser scan, approach 2 photogrammetry. 

The 123 product and recap photo seem to use approach 2 while the blk360 uses approach 1.

why not use photogrammetry as the starting point and refine the information with the laser data? Or is that what it already does?

[ryan.frenz]

I wouldn't say the approaches are competing - more complimentary actually.  However I'm not sure that applying photogrammetry here would be worth it.  These approaches require a lot of compute to reconstruct the 3D 'scene' behind each photo.  This 3D info is free with a full-dome laser scanning system like the BLK.

The registration step is a bit more interesting.  Here the photos could help in some edge cases where the scene's features are only visual, not geometric (the long straight hallway problem).  However, in the normal case, you'd always prefer the laser since it has no light/color/exposure dependencies.  And certainly there are tons of spaces where features are only geometric (especially inside buildings).

Just my 2 cents.

-Ryan