Three-dimensional model representation method and system for expressing geometric details and complex topology

Abstract

The invention provides a three-dimensional model representation method and system for expressing geometric details and complex topology, and the method comprises the steps: inputting a group of three-dimensional geometric models, with semantic tags, of the same type, and calculating a bounding box for each part of the model; registering the bounding box to a corresponding component to obtain a corresponding deformation gradient, and further obtaining a deformation gradient vector of the component; obtaining a distribution vector of the component deformation by using the component deformation gradient vector through a component variation auto-encoder; analyzing the global structure of the model by taking the support relationship as a main part, and constructing a representation vector of each part; connecting representation vectors of all components of one model in series to serve as input, and changing the global structure and geometric details of the auto-encoder joint encoding modelthrough a structural component; randomly generating a new model through the trained structured component variational auto-encoder, or interpolating between the two models to generate the new model; and performing global structure optimization on the generation model under structural constraints and stable support constraints.

Description

technical field [0001] The invention relates to the fields of computer graphics and deep learning, and in particular to a method and system for generating a three-dimensional model representation capable of expressing geometric details and complex topological structures. Background technique [0002] With the development of virtual reality technology, the demand for 3D geometric models in the industry is increasing day by day. Traditional geometric modeling methods require users to have corresponding professional skills and use professional software such as AutoDesk for modeling. This increases the cost of 3D modeling and limits the pool of people who can do geometric modeling, which in turn limits the ability of users to add models in virtual reality and the variety of 3D printing. The present invention mainly relates to the field of representation learning of 3D geometric models. Recent research attempts to use deep learning methods to learn representation of 3D geometric...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to solve the problem that the above-mentioned prior art cannot accurately describe the details of the geometric model, and propose a high-quality three-dimensional model representation and generation method, based on the method of bounding box deformation to describe the details of the grid model, using the component variable Descriptive autoencoder encoding deformation information and two-stage variational autoencoder architecture to describe the geometry and topology of 3D geometric models

Images