Do a Gage R&R on the types of features you will be measuring. A good engineer is not going to argue real world data tests.
As far as algorithms to use, it depends mainly on the application. With rubber parts, you will have a lot of form problems that will not affect functionality of the part but will affect the reproducibility of the measurement results. I would opt for a least squares solution to 'circles' of this form with ALOT of data density. A practical way to test if you are getting enough points (and you don't have the money for a scanning system like Zeiss' VAST) is to output the form and keep increasing the points till diminishing returns of the form. At this point, you will have collected enough points to roughly capture the form.
There are more scientific approaches, but this is practical and will net you pretty good results. Hope it helps. tas
Timothy A. Schroeder
Supervisor- CMMs, Gearlab, and Gage Inspection
Rolls-Royce Corporation
P.O. Box 420, Speed Code R29A
Indianapolis, In 46206
Tel: (317)230-5914