Research on shape factor based on Gaussian radial basis function

Abstract

Surface reconstruction technology needs to fit the collected point cloud data into an expression with a complete base function. How to improve the fitting accuracy of surface reconstruction is the focus of this research. In this paper, aiming at the surface fitting model based on Gaussian radial basis function(RBF), the influence of shape factor selection on surface fitting accuracy is studied, and a new method for calculating RBF shape factor is proposed. The shape factor is connected with the size of the surface to be fitted, which strengthens the relationship among the kernels. Taking the given point cloud data as an example, the simulation experiments were carried out, and the fitting accuracy of different shape factors was compared, and the curved surfaces of different shapes were fitted. The results show that the accuracy of surface fitting can be improved by choosing shape factor combined with the size of fitting area; surface fitting based on RBF is suitable for the surface with small curvature and gentle surface; RBF can be used for the description of non-rotational symmetric surface, but the fitting accuracy will decrease.