Machine Learning Based Complex Surface Feature Extraction And Segmentation Method And Its Applications

The segmentation of 3D model is a key step in 3D model processing.The segmentation of 3D model plays an important role in 3D model retrieval,geometric compression transmission and 3D model simplification.3D model segmentation method based on curvature information is one of the most important segmentation methods in 3D model segmentation.Based on the summarization and analysis of the research status of the 3D model feature extraction and segmentation,the 3D model segmentation method based on curvature information has disadvantages like low efficiency and no guarantee of accuracy,and so on.According to the research results a 3D model segmentation system with complex surface is developed,which is validated in the segmentation of tooth model.The main contents of this thesis include:The first chapter introduces the domestic and international research on feature extraction and segmentation of 3D model based on machine learning.Including 3D model feature extraction technology,based on machine learning threshold estimation technology and 3D model segmentation technology.In the meantime,analyzes the research progress and existing problems of the above-mentioned technology,and puts forward the content of this paper and analyzes its research significance of the text structure.In the second chapter,the feature description of Gaussian curvature density histogram of 3D vertex in complex surface is proposed.This feature description is of great help to the acquisition of Gaussian curvature threshold in 3D model.Based on the feature description of the Gaussian curvature density histogram,the design of 3D model features is proposed.At the same time,the experiment is designed by using the related tools,and the Gaussian curvature density histogram of the 3D model is extracted.The correctness of the Gaussian curvature density feature extraction method is verified.In the third chapter,a Gaussian curvature threshold estimation method for complex surfaces based on machine learning is proposed.According to the features of complex samples and their threshold of Gaussian curvature,the regression equation is trained.The parameters of the regression equation are improved and corrected by the complex surface models.The reliability of the regression equation is also improved.The Gaussian curvature threshold for complex surface model can be solved according to the complex model feature,and the curvature-based in the process of complex surface segmentation,the Gaussian curvature threshold is difficult to calculate.In the fourth chapter,a 3D model segmentation method based on Gaussian curvature threshold estimation is proposed.Based on the Gaussian curvature threshold estimation method proposed in Chapter 3,a regression model between the complex surface model feature and the corresponding Gaussian curvature threshold is established,then the Gaussian curvature threshold of the model to be segmented is solved by using the regression model.And the segmentation efficiency of the complex surface model is improved.By segmenting the 3D model,the complex surface segmentation algorithm based on machine learning and curvature information is designed and validated.In the fifth chapter,we develop a 3D surface segmentation system for complex surfaces.We implement a feature extraction module for 3D surface segmentation of complex surfaces,a Gaussian curvature threshold module based on machine learning estimation,and a 3D model segmentation module based on machine learning and curvature information.And the system is applied to the tooth 3D model segmentation.The sixth chapter summarizes the full text and summarizes the shortcomings of this paper,and makes a prospect for the future research work.