Research And Application On Example-based Digital Geometry Modeling System

The technique of modeling by example was proposed by researchers at Princeton University. It can create new models through deforming and assembling existing models or model components. Compared to traditional modeling technology which needs to be operated by a small set of trained experts with complex commands, example-based modeling technology aims at general public, allowing them to create detailed geometric models quickly. Novice users can deform and combine existing models or parts of models from examples in the database with simple and intuitive interactions such as sketching.Our goal is to build an example-based digital geometry modeling system. First of all, we analyze the technology problems and system problems that need to be solved in case of construction of modeling system. Then we make an intensive study of the ways to figure out the system problems and to design a hierarchical framework to well integrate existing technique modules that relate to example-based modeling. The contents of this thesis are mainly embodied in the following aspects:(1) Design and implement of digital geometry modeling system. Based on the analysis of related problems, we modularize the system, define the functions of each module, clarify the relations between modules and use a hierarchical architecture to organize each module.(2) Mesh transformation technology. This part of work aims at the data adaptive problem of example-based modeling system. Considering the constraints that the modeling system put on the mesh transformation algorithm, we proposed a mesh transform process that suits for example-based modeling system combining surface-based and volumetric mesh transform algorithm. In the process, we not only remove the artifacts in the input model, but also keep the input and output models alike both in global and local features.(3) 3D shape retrieval. Aiming at the data acquisition problem of example-based modeling system, we compare existing similarity measure methods for models on the background of example-based modeling system, and select four model features fitting for example-based modeling system. Then, we use several criteria to evaluate their retrieval performance under two retrieval strategies separately-overall retrieval and parts retrieval. The result turns out that parts retrieval outperforms overall retrieval in both single- and combined-feature and integration of these two retrieval strategies with suitable weight can reaches a higher retrieval performance.