This paper presents a general method for automatic surface fitting from
scattered range data and describes the implementation of three methods for fitting
surfaces: linear, quadratic and cubic. It uses a modified 2D least squares method to
fitting, reconstructing and modeling several surfaces and statistical criteria to
compare the three approaches. The comparison is performed using a
mathematically defined data as real data obtained from the proposed models.
The method can be used in a variety of applications such as reverse engineering,
automatic generating of a CAD model, etc, and it has proven to be effective as
demonstrated by a number of examples using real data from mathematical
functions ( sine, cosine, exponential and cubic). By applying the proposed
surface fitting model the standard deviation was found to be (0.04-0.26), (0.02-
0.07) and (0.0-0.12) mm for linear, quadratic and cubic fitting models