Ellipse detection with hard c-regression models and random initializations

Shell clustering methods partition data sets into several shell-shape clusters by extracting local circles or ellipses as prototypes of clusters. This paper proposes hard c regression models (HCRMs) for shell clustering. The procedure is a defuzzified switching regression models. HCRMs successfully detect ellipses by using random initializations. We report the performance using 20 data sets each of which consists of two ellipses. The detection time on average is 14 milliseconds on DELL PRECISION T5400 3.16GHz.