Research On Triangle Mesh Models Segmentation,Evaluation, And Their Applications

Automatic segmentation of3D surface meshes into functional parts is a fundamental problem in computer graphics. A partial decomposition not only provides semantic information about the underlying object, but also can be used to guide several types of mesh processing algorithms, including reconstruction, compression, texture mapping, parameterization, texture mapping, mesh deformation, model retrieval etc. Investigation on3D mesh segmentation research literature exhibits that it is still one of the most active research areas in graphics after more than two decades’ extensive and profound research.So far many automatic mesh segmentation algorithms have been developed over the last several years by domestic and foreign researchers。Unfortunately, different algorithms have its own applicable scope and limitations. Selecting a unsuitable algorithm or geometric characteristics will serious influence the efficiency and the accuracy of segmentation. It has been an arduous and time-consuming task to select an appropriate algorithm from the existing algorithms for a specific application, which also attracts the more researchers’ attention。Compared with the mesh segmentation technology with the mature image segmentation and evaluation technology, there has been less work on mesh segmentation quality and evaluation on the benchmark. What we can review and analyze from the existing mesh segmentation criterion, evaluation methods and similarity measurement is that as human experts’ segmentation benchmark exists great subjectivity and randomness, similar the evaluation criterion for evaluation of the same segmentation results has distinct differences, which means quantitative evaluation criterion of3D mesh segmentation still needs further exploration and research. Based on this conclusion, this paper established the evaluation of segmentation on2D manifold meshes and its application as the research content.After a description on the research background and significance, a review on domestic and international research status, the dissertation has executed integrated and extensive research on the planned goal, and conducted enormous base and verification experiments. The achievements can be summarized as follows:(1) Appended the concave to optimize the lowest points of water collecting basin, and also in flooding process, a new strategy is proposed by considering both local and global features. Tests of Princeton data set show that the algorithm in CD, CE, HM, index error of RI has a relatively good results. The key point of3D model retrieval is the extraction of shape features and the computing the similarity between models. However, most of the current methods consider only global information of models without considering the partial structure,and thus resulting in low accuracy of retrieval rate.3D mesh models searching based on segmentation which firstly divide3D mesh models into meaningful parts, then the local information and features of each block are calculated based on the topological relation of blocks of models to map the3D model a feature topology graph,then creat a spanning tree from the feature topology graph. and the last step is calculating the similarity between the trees of meshs.. Experiment shows that the results which based on the significant segmentation become more in line with people’s cognitive requirements.(2)A fast query of adjacency triangle algorithm in triangle mesh is proposed. Dealing with triangular manifold in the grid often needs to query some triangles’ adjacent area and number of their adjacent triangles. But in off or obj format files which build the mesh only provided the information of each vertex of each triangle. It is not possible to obtain information related to the triangle and the adjacent triangles’ directly from the file (at least one common vertex). Violence enumeration method which has already existed is confronting high time complexity. In order to solve this problem, an algorithm is proposed to fast query triangle adjacency triangle lists.(3) Extension of the existing evaluation system of Princeton which presents a model of fuzzy evaluation of segmentation quality, a method for the computation of calculating method of fuzzy membership degree, and gives a calculation of the two set of fuzzy correlation method. It can be verified that the segmentation and evaluation method based on fuzzy theory can be better with human cognitive consistency. Membership function of the fuzzy set reflects the element belonging level to the set. This part of the study discuss about fuzzy operation problems, calculation of the commonly used T Paradigm and T co paradigm operator, the extension of the existing evaluation system, presents a novel method for fuzzy evaluation of segmentation results. Presents a calculation method of fuzzy membership degree set concerning the correlation, and gives a calculation program related degree.(4) To explore the innovative application of mesh segmentation. The zero watermark can extract the structural feature of the model without changing the appearance which is a fundamental solution to the contradiction between strength and transparency. Therefore, searching for unique identifying characteristics is the key point to the problem. Aiming at this problem, this paper starts from the model segmentation, a scheme is proposed by grouping construction and recombination,thus create a3D mesh watermarking. For the shape diameter function of each triangular is selected as the structure feature which is robust ultimately not only geometric rotatation, translation,scaling but also on the mesh simplification.