Abstract

This paper reports a novel 3D surface modeling method based on nonlinear auto-regressive moving average with exogenous input (NARMAX) in time domain - 3D Surface modeling based on spatial neighbor points coupling (SMSNPC) for scattered data points. The 3D model of complex surface can be reconstructed through the mutual coupling of neighbor points. The algorithm of adjustment for model parameters is introduced in this paper based on ANN. To demonstrate the efficient applicability of the developed prototype, two groups of point cloud data are presented. One collected by the robot measure system is used to validate the reconstruction result and the other is used to compare with conditional non-coupling ANN modeling in reconstruction precision. From the results, the new approach presented in this paper validates the expectation that the surface modeling based on spatial neighbor points can match the real surface satisfactorily and it is particularly suitable for modeling of complex surface which cannot be finished by non-coupling ANN modeling. And from the comparative studies it is shown the reconstruction precision of SMSNPC is much higher.

Key words: Surface modeling, scattered points, coupling, ANN, SMSNPC.