Planar Curve Fitting by Simulated Annealing with Feature Points Determination

摘要

The determination of an approximation curve from a given sequence of points is an important task in CAD. However, the determination of the minimum number of required parameters to represent the curve is an open issue. This work proposes a semi-automatic method to determine a piecewise curve that approximates the given sequence of sampled points, in which a manual step is done to determine the number of required parameters and an automatic step is done to determine the curve parameters. Feature points are manually chosen within a simplified sequence of points of the sampled sequence. These feature points determine the number of cubic Bézier curves segments, and a simulated annealing algorithm is used to minimize the discrepancy between the given sequence of sampled points and the piecewise curve by adjusting the curve parameters. The curve length is used as regularization to avoid the overfitting problem.