B-Spline Curve Fitting onto Measured Point Data under Consideration of Curvature.

Constructing a CAD system for producing CAD models from a master model is important to style design. This paper describes curve fitting onto the measured point data of a master model. The point data were measured with a non-contact 3-D digitizing sensor, and a 4-degree non-uniform B-spline curve was fitted onto them. In the process was used a non-uniform knot vector determination method which took account of curvature as well as method of calculating curve fitting errors in the normal direction. Comparing the uniform B-spline curve with a curve obtained in the new fitting methods was simulated with a computer. So was the influence of the measured point data containing measurement errors. Control points were calculated by B-spline curve inverse transformation based on the least-square method. Then inspection experiments were performed by use of a master model having freeform surfaces. As a result of the experiments, it can be concluded that the new fitting methods can help reduce the 404 measured point data in the experiments to 23 control points and 27 knots, and that the B-spline curve calculated in the new methods can he fitted onto the measured point data with high accuracy in the whole range.