Three-dimensional grid model retrieval method based on geometry image

Abstract

The invention discloses a three-dimensional grid model retrieval method based on a geometry image. The method comprises the following steps that: mapping a three-dimensional grid model to a sphericalsurface through an area preserving parameterization method to obtain a parameterized spherical three-dimensional grid; then, mapping the obtained spherical three-dimensional grid to an octahedron, andshearing and paving along the edge of the octahedron to obtain a two-dimensional plane; utilizing three different types of geometrical characteristics of the three-dimensional grid model to code eachpixel on the two-dimensional plane to obtain a geometry image; adopting a standard convolutional neural network structure, taking the geometry image as input, obtaining the characteristics of the geometry image through iterative training learning, and obtaining the characteristics of the corresponding three-dimensional grid model through maximum value mapping; and according to the obtained characteristics of the three-dimensional grid model, calculating a similarity between the three-dimensional grid model to be retrieved and rest three-dimensional grid models in a database, and outputting aretrieval result according to a similarity calculation result.

Description

technical field [0001] The invention relates to the fields of computer vision and computer graphics, in particular to a three-dimensional grid model retrieval method based on geometric images. Background technique [0002] With the advent of the information age, the 3D mesh model, as a new multimedia data, has been widely used in the fields of computer graphics and computer vision. At the same time, 3D modeling technology is also developing continuously. The emergence of various 3D sensors makes it easier to obtain 3D mesh models, such as Microsoft Kinect, Google ProjectTango, etc. Nowadays, there are many rich 3D mesh model libraries, and they are easy to obtain online. How to manage and analyze them, 3D mesh model retrieval technology is one of the important methods. [0003] The retrieval methods of 3D mesh models are mainly divided into text-based retrieval and content-based retrieval. Text-based 3D grid model retrieval first semantically annotates the 3D grid model, a...

AI Technical Summary

Problems solved by technology

[0007] For 3D data that relies on voxel or point cloud representation, using a 3D neural network for training and learning suffers from very high computational complexity, disordered data, and lack of finer geometry.

In order to unnecessarily adjust the convolutional neural network architecture to adapt to the surface convolution operation, it is necessary to convert the 3D mesh model into the planar structure required by the convolutional neural network, such as projecting the 3D mesh model from multiple angles to obtain multiple views. , but this way will lose a lot of geometric information

Images