Fitting conics of specific types to data (Q1826713)

The author presents an algorithm for fitting to data a conic of a type specified in advance, by minimizing a quadratic form subject to two simultaneous quadratic equality constraints. An orthogonal change of variables is used to transform this problem into a least-squares problem with one quadratic constraint. The resulting best fitting conic remains invariant under Euclidean transformations. The algorithm is applied to one example arising in the study of sundials, and comparisons are provided with algorithms and test cases from the literature.