Research On Mesh Segmentation Algorithm Based On Persistent Clustering And Multi-diffusion Signature

In recent years,with the rapid development of digital scanning technology,3D animation and the Internet,a new research branch has been formed on the processing,analysis and representation of 3D models,integrated is the 3D model editing.The main research contents include mesh denoising,reverse engineering,mesh segmentation,shape matching,texture mapping,model repairing,and geometric deformation.Because of the complexity of the 3D model itself and the difference of the segmentation results that users want to get,there has not been a unified algorithm for mesh segmentation research for a long time,and there will be more or less shortcomings in the existing methods for model segmentation.In order to adapt the segmentation results to multiple models and improve the segmentation speed,this paper proposes a multi-signature mesh segmentation algorithm based on persistent clustering.The main work is as follows:(1)The paper studies and analyzes the commonly used segmentation algorithms,mainly introduces the application and characteristics of several types of segmentation algorithms,summarizes the representation methods of 3D mesh feature points,and performs existing wave kernel signatures and thermonuclear signatures.The advantages and disadvantages of wave kernel signature and thermonuclear signature are compared in detail.In addition,this paper classifies and elaborates the clusters with more applications.(2)A segmentation algorithm based on persistent clustering wave kernel signature is proposed.Firstly,the wave kernel feature function is calculated,and then the persistence graph is generated by persistent clustering.Finally,the number of segmentation components is selected to generate segmentation results.Experiments were conducted under the segmentation standard provided by Princeton University.The overall segmentation algorithm of this paper is superior to most unsupervised segmentation.(3)Based on the existing signature features,persistent clustering segmentation algorithm based on fusion-diffusion signature's are proposed.The segmentation standard test was carried out on the experimental results,and a good segmentation result was obtained.In order to enhance the integrity of the experiment,this paper also explores the mesh segmentation in combination with extreme learning machine.