Poor surface fit in OCC5.0

I have 200 data points that fit (were computed from) a spiral patten on the rounded surface of a sphere. I am using mfc sample 11_plate to see how well OCC5.0 can fit a surface to those data points. The sphere's diameter is 10.0. The data points start at 0,0,0 and go out to 4,0,-8. OCC comes up with a surface that doesn't fit the points well at all. The surface has many folds near 0,0,0 that are not justified by the data and spreads out well beyond the data. Is there any reason for the poor surface fit? Does the order(placement) of the data points matter to the surface fit? Is there a limitation to the number of data points? How can I improve the surface fit to the data?
Thanks very much for any advice.

Francois Lauzon's picture

I don't think the plate algorithm could fit a closed surface. You will have to split your points into region that don't form a close surface.
- There is no order for the data points.
- There is a limitation to the number of constraint after which the fitting algorithm might become unstable, we found it to be around 1500-2000 points, depending on the shape, it might be able to compute or not, but it will that time (sometime around 20 minutes).
- One way to improve the surface fitting is to give an initial surface to the approximation algorithm (by default, Plate is building a plane), or to play with the way Plate creates it's default surface, for all that you should have a look at the documentation in the cdl files.
- There is also a better algorithm called PlateFE, which is much faster and can take a lot of points (they say near millions of points), but you have to buy it from OCC.

Good Luck,