Parameter optimization in approximating curves and surfaces to measurement data (Q1183519)

The authors investigate the process of parameter optimization. They propose a parameter optimization method based on the nonlinear least- squares technique due to Levenberg-Marquardt (the error expression is derived using cubic \(B\)-spline basis functions, this expression is then minimized using the parameter values at each point as variables). Unlike usual methods, least squares fitting of the curve or surface is performed only once, after the parameters are optimized, instead of repetitive fitting. The authors compare their optimization process with methods published by \textit{J. Hoschek} [ibid. 5, No. 1, 33-40 (1988; Zbl 0647.65007)], \textit{J. Hoschek, F.-J. Schneider} and \textit{P. Wassum} [ibid. 6, No. 4, 293-306 (1989; Zbl 0682.65005) and \textit{D. F. Rogers} and \textit{N. G. Fog} [Computer-Aided Des. 21, No. 10, 641-648 (1989; Zbl 0687.65008). Finally examples are provided for the approximation of curve and surface offsets following the same approach as is described by \textit{Chen} and \textit{Ravani} [J. Mechanisms, Transmissions and Automation in Design 3, 133-142 (1987)].