FITTING CIRCULAR ARCS TO MEASURED DATA
Les A. Piegl, Wayne Tiller · International Journal of Shape Modeling · 2002
Algorithms for fitting circular arcs and straight line segments to measured data are presented. Algebraic as well as geometric methods are discussed leading to efficient techniques for arc and line fitting. A thorough empirical analysis reveals that the best circle fit is via algebraic minimization followed by distance minimization, whereas the best line fit is obtained by a simple geometric method of direction vector averaging. An automatic fitter is also presented that determines whether a line or a circle fit is optimal, computes the appropriate entities, and clips the geometry to obtain the best NURBS circle or line fit.