A dynamic neural network model for accelerating preliminary parameterization of 3D triangular mesh surfaces

Abstract

This study proposes an effective and fast preliminary mapping algorithm for 3D triangular mesh surfaces. The proposed method exploits barycentric mapping theory and dynamic neural network for computing parametric coordinates corresponding to vertices of 3D triangular mesh. The dynamic network model iteratively moves internal nodes in 2D parametric space until they convergently reach an equilibrium state. The method effectively computes parametric space coordinates of large meshes (having more than 1.5 K vertices) in less time compared to the traditional method using inverse matrix calculation. The proposed method is tested on many surfaces of varying size, and experimental results prove its efficiency and efficacy.