Abstract
A new method of fitting implicit conic to plane scattered data points is presented in this paper, which is based on minimizing the sum of squared point-to-curve algebraic distance. At first, the specific ellipse, hyperbola and parabola are fitted to the data points respectively, then the final fitting conic is produced by combining the above three specific conics and adding certain weights to the coefficients of them. By this method, the fitting conic not only preserves the original curve shapes if the conic data comes from the basic quadratic curve, but also improves the fitting effects for the general data points.