Surface classification: hypothesis testing and parameter estimation

A 3D surface classification method based upon the quadric surface model is described. This technique does not require the points from the surface to lie on a grid. A sample of surface points is classified as planar or nonplanar through two hypothesis testa If the sample is nonplanar, curvature features are evaluated at each point to classify the sample as spherical, cylindrical, or conical. A nonlinear optimization technique is then used to refine the parameters (e.g. rdius, orientation) of the resulting surface type.