Study On Part Replacement And Mesh Model Generation Based On Samples

3D model is a widely utilized digital media form in CAD,CG and CV,etc.With theincreasing model number and the rapid development of digital geometry processing,it has become an important issue of mesh model processing to analyze,reuse the existing mesh models and modify it according to the new design requirements.Based on 3D mesh samples,this master thesis investigates a model generation methodology by part matching and replacing.A part matching method between two mesh models is studied based on hierarchical structures.Firstly,shape analysis is implemented for mesh models,including model normalization,contact analysis and symmetry detection.Then,a hierarchical structure representation that includes the contact relation and parent-child relationship could be designed.And the hierarchical structures of mesh models could be constructed according to the coordinate position information of their parts.According to the nearest neighbor principle,the nodes in each level of the two hierarchical structures are adjusted by allocation,deletion and merging node operations.After that a one-to-one part matching is established from coarse to fine for them.Another part matching method is studied based on statistical random point distance distribution.The geometrical features of the modes are characterized by statistical distribution.The Euclidean distance of any two uniform sampling points on the model surface is calculated and the number of distances in each distance interval is counted,after which a histogram is obtained for the statistical distribution of each part,and then part matching is carried out by comparing the similarity between the histograms.For a case where the statistical distribution histogram of a part is similar to multiple parts',the gravity center distances between parts are combined to match parts.Novel models are generated through part replacing and contact processing after part matching.In order to make the intermediate models generated gradually transition from the original model to the target model,a method of stepwise replacing parts is used.The contact processing between parts of new models is performed based on the adjacency relation.A translation transformation for parts of the models reorganized is used and the corresponding parts are translated to the target location,so that contact positions are reasonable.A model generation system has been developed based on sample part matching andreplacing.The experimental results show that the model generation algorithm cloud be quickly generate a series of novel models and generated models are different from the original models in appearance,but similar in symmetry and adjacency structure.