Ah! I see.
Good algorithms to fit NURBS surfaces over a facet model are pretty 'high end'.
My practical experience is limited to manual surface creation using the facet
models as (indirect) reference. (Xsec can be used to create visual references,
Rhino, if available, can create cross section curves for import.)
Rhino has a MeshToNURBS command but all it does is convert each facet (vertex
list and orientation vector) to a NURBS surface (CVs, knots, bounding curves)
definition. It's still 'faceted'. There is (/ was?) an add-on called SMURF
that fits surfaces over groups of facets. I've never seen anything to gauge
it's capabilities.
Direct Dimensions (dirdim.com) has some of the best samples on their examples
page I've seen (and that is about the extent of my knowledge of the subject).
The human head example was done in (? exported from) Geomagic Studio. The
machined cover was exported from Solidworks. What came before export I don't
know.
Looking at Help: ReStyle, with a Reverse Engineering license, has functions
for the purpose. (Think I've read they're Geomagic based.)
About the only other Pro/E option (a manual surfacing variation) that I can
think of to try (haven't but it sounds like an interesting experiment) would
be to strip the vertex coordinates from the *.obj and read them into a
Scan-Tools Independent Geometry feature as High Density Raw Data for Scan Curve
creation.
