3D Feature Representation Based On UV Unfolding And Application Research

UV unfolding is a technique used in the field of computer graphics for texture mapping on the surface of 3D models,mainly through the deformation of the model surface mesh,to establish a one-to-one mapping relationship between 3D coordinates to image coordinates.Since the surface of objects,especially non-rigid models(e.g.human body),is not invariant to spatial coordinates,while its 2D coordinates relative to the surface are generally fixed,therefore,3D models can be transformed into 2D planes for research.On the other hand,image representation of 3D models has certain application value and practical significance for object recognition based on deep learning.To address this issue,this paper focuses on how to transform 3D models into images and realize the dimensionality reduction of 3D models.Feature extraction and recognition are performed based on the image representation of the 3D model to verify that the 3D features can be stored on the image after downscaling.UV unfolding of 3D models makes the huge 3D model information reduced to 2D information,which has great significance in 3D feature learning and 3D recognition.In this paper,we map 3D surface to UV space through UV unfolding for 3D feature representation research,which mainly includes the following aspects.1)3D features based on point clouds.Firstly,the point cloud features are described in detail and used to align the point cloud data and compare the advantages and disadvantages of each alignment algorithm.2)3D features based on UV expansion.This topic focuses on the two-dimensional image representation of three-dimensional features.By mapping the three-dimensional model to twodimensional UV coordinates,the huge point cloud data is transformed into a simple twodimensional image.Firstly,the 3D model is divided into closed and non-closed models for UV unfolding,secondly,the 3D point cloud is projected onto the 2D plane after PCA processing to generate depth images,and finally,the 3D meshing is studied and unfolded into UV coordinates by different algorithms,and the UV space is used to store all the data in the 3D dataset,so that the3 D data and images correspond one to one.3)UV unfolding feature extraction and recognition.After UV unfolding,the 3D model is extracted and identified with features.The experimental results show that the 3D feature representation method based on UV unfolding proposed in this paper can unfold the 3D model onto the image and obtain the ideal experimental results.This paper can give priority to ensure that the surface of the 3D model is kept intact on the dimensional plane,and the surface features of the 3D model can be extracted by indexing the 2D coordinates,and finally the pattern recognition experiment of the 3D model is carried out based on the UV image.