Robust fitting of ellipses with heuristics

In this paper we solve the problem of robust fitting of ellipses to a set of points under condition of very high noise (when more than 50% of the points are outliers), using the sum of the orthogonal distances (instead of the sum of the squares of the distances) from the given points to the fitted ellipse. The fitting problem now is nonlinear and is solved using Differential Evolution. We demonstrate by simulations that one or several ellipses can be extracted from the data set of points. The proposed method is compared with the least median of squares solved with RANSAC.