Study On Multiresolusion Analysis-based Feature Extraction Of The Triangular Mesh

With the rapid development of modern science and technology, ReverseEngineering is widely used in the developmengt of the new product and theachievemengt in the modern manufacturing industry. In this, the feature informationreflects the design intention and characteristic of the product,which is momentoussignificance in Reverse Engineering.However,the coexistence of multi-scale features isa common phenomenon in the design of complex products.The key problems of thetroubled designer are the local feature’s scale analysis,the feature boundary’s partitionand the feature’s continuity and so on. In the background of the ReverseEngineering,based on the triangular mesh,the method of the feature extraction and themulti-scale feature analysis are studied in this thesis. Then,combined with the waveletanalysis,the curvature properties of the triangular mesh are studied too.The mainresearch contents are as follows:1.The method of the feature extraction based on the volume intergal isproposed.Aiming at the property of the integral invariant irrelevant to the meshes’topology and the inner relationship between the volume integral invariant and themean-curvature, used the global4D Shepard surface to estimate the curvature of eachpoint in the triangular mesh, then the volume integral invariant was calculated by thegiven formula, and the feature points were extracted through the K-Means clusteringalgorithm.This method is only related to the numbers and the operations of the pointscloud,and the experiments show that this algorithm has higher computing efficiency,andexpressed the boundary feature of the clouds more precisely.2. The method of the muli-scale analysis of the feature scale is proposed.First ofall,the local scale factor is defined to analysis the geometric properties of the triangularmesh in the n-ring neighborhood.Then,using the orthographic projection methodprojects the mesh’s vertex. Then computer the vertex’s scale facator value.Then,usingthe Histogram equalization algorithm makes the Histogram approximately uniformdistribution.Last,by the graph-based segmetation method,the multi-scale featureanalysis is realized.The experiment shows that this method can not only reflect the scaleof the feature better,but maintain the geometric characteristics of the surface better.3.Combined with the wavelet technology(multi-resolution property) and the local scalefactor,analyse the triangular mesh’s curvature property. Exploiting the detail-removed wavelet reconstruction method,the curvature distribution of the triangular mesh surfacecan be acquired.In theory,this method can well describe the curvature of the triangularmesh surface.