An iterative procedure for robust circle fitting

The problem of fitting circles and circular arcs to observed points arises in many areas of science. However, the fitting results by using most geometric and algebraic methods are not usually acceptable in the presence of outliers. An iterative procedure for robust circle fitting is proposed. During the iteration, Taubin's method is employed to obtain the center and radius. And then the geometric distances from the data points to the circle are computed, with which outliers are identified and removed. Numerical examples demonstrate that the proposed iterative procedure can alleviate the corrupted effect of outliers on the circle parameter estimates.