Converting RTI data to Point Cloud Data

[3dguy]

Has anyone developed a program that will fabricate a point cloud from the data in a PTM file?

[cdschroer]

There is discussion on the forums about producing 3D from RTI data.  You can find the discussion here:

http://forums.culturalheritageimaging.org/index.php?/topic/64-3d-conversion/?hl=%2Bnormal+%2Bfield+%2Bconvert&do=findComment&comment=68

 

RTI's do not produce point clouds or 3D surfaces.  They produce reflectance properties and normal fields and illumination independent color.  While there are ways to convert a normal field to a 3D surface, there are issues (see the above mentioned thread)  Remember that with RTI there is only a single camera position.  To properly produce low uncertainty 3D points you need multiple look angles.  If point clouds are your goal, we recommend using photogrammetry.

 

Carla

[3dguy]

Thanks Carla, I'll digest those links and probably get back with you. I realize that RTI files are "just" a reflectance/pixel file but your point in the other post, the accurate high resolution/high frequency surface information is what we're trying to capture. You mention " There is a small error in the calculus........". Is that in the normal data?

 

Perhaps, using the RTI point cloud in conjunction with photogrammetry may correct the distortion. In our project, which is a small wooden relief lacquered panel neither photogrammetry or laser scans have given us the resolution we want. The RTI data does.

[cdschroer]

You might want to review this thread for a discussion on combining 3D data and Normal data:

http://forums.culturalheritageimaging.org/index.php?/topic/303-efficiently-combining-positions-and-normals-for-precise-3d-geometry/

 

The error is not in the RTI data, the error happens due to an uncertainty n the calculus when you convert a normal field to a 3D surface.  The uncertainty accumulates and it causes the resulting surface to warp.

 

We are getting great results on very detailed surfaces with photogrammetry. If this is your goal, you really have to plan out the photography capture and follow steps to ensure low uncertainty, and very high quality camera calibration, as well as multiple look angles at every point on the surface.  These steps can radically reduce  the high frequency "noise" that often occurs in 3D data.  (warning: shameless plug coming)  Come take our photogrammetry training class, and we will teach you all the concepts and practical planning, imaging, and processing steps to reduce error and get measurable results.

 

Carla