A parameterized geometric fitting method for ellipse

Abstract This study presents an accurate and robust method for fitting noisy and occlusion elliptic data. The nonlinear issue of ellipse fitting is interpreted as a set of Levenberg–Marquardt iterations (LMIs) by minimizing the geometric distance. For each iteration of the geometric fitting, the representations of the parameterized ellipses are mapped to the geometric error distances, which are implemented latently by an orthogonal angle segmentation of the ellipses. Moreover, dimension reduction is utilized by the LMIs to avoid misconvergence and expensive computations. The method based on two recent representative algorithms is verified by simulation and real-world experiments. The results suggest that the saliency capability of the new method to fit ellipse-specific profiles with severe noise and occlusion, which is better than or equal to those of the reference approaches, has potential applications in quality monitoring, three-dimensional reconstruction, and instrument calibration.