三维网格模型聚类分割
摘 要:
三维网格模型在计算机图形学、三维动画、计算机辅助设计等方面有广泛应用。通过三维激光扫描仪的数据采集，将物体进行数字化处理，构造、生成相应的三维网格模型，利用聚类分割技术对生成的三维网格模型进行重新划分、提取、组合，根据需求构造出新模型、方便新课题的研究已成为新趋势。
本文的主要目的在于学习、探索、研究、测试三维网格模型的聚类分割方法，并对美国普林斯顿大学的三维网格模型库中的模型进行遴选、测试。
在三维网格模型的数据分析处理过程中，可以根据三维网格模型上任意点到该模型上其他任意点的距离对属于模型的点信息进行聚类，也可以根据三维网格模型上任意面片到该模型上其他任意面片的距离对属于模型的面片信息进行聚类。由于在实验过程中考虑到采用任意点到其他任意点的距离聚类虽然实验结果更准确，但在真实感绘制环节中存在判断点与面片的归属、面片重新聚类、渲染丢失面片等问题，所以本文采用了根据三维网格模型上任意面片到该模型上其他任意面片的距离对属于模型的面片信息进行聚类的方法，重点对8个模型进行了聚类分割测试。
本文采用了K均值聚类算法完成了三维网格模型的近似有意义的分割。首先在visual studio 2005平台下采用了Floyd算法计算了三维网格模型任意面片到其他任意面片的距离，并根据得到的计算结果在MATLAB平台下完成了K-means算法的三维网格模型的近似有意义的分割。本文对dijkstra和Floyd算法进行了分析比较。完成了基于MeshLab平台下的分割结果真实感绘制，对分割结果及渲染效果做了具体说明。
关键词:
三维网格模型 聚类分割 Floyd算法 真实感绘制
Abstract：
Three-dimensional grid model has wide application in the area of computer graphics,3D animation,computer-aided design and so on.Three-dimensional laser scanner makes data acquisition available ,which means by digital processing the object and the structure of model,it can generate three-dimensional grid model.Cluster segmentation of the mesh model has a range of applications,such as three-dimensional re-divided,extraction,combination of different pars of the previous model.It's a new trend to construct a new model on demand model to facilitate new research by using this method.
The main purpose of this paper is to study、explore、do research and test three-dimensional grid models based on cluster segmentation.Princeton University's three-dimensional grid model base offers more choices to model selection and testing.
In the data analysis process of three-dimensional grid model,two basic ideas are considerable.One method to handle the issue is :calculate distances between arbitrary point in the model to any other points in the same model,then do the job of cluser segmentation.Another method to solve the problem is:calculate distances between arbitrary patch in the model to any other patches in the same model,then accomplished segmentation for the model.Though more accurate results are expected by using the method one，after considering the issues of hardly judging the attribution of points to patches,re-clustering of patches and losing patches while rendering,I prefer method two to solve the problem.In the test ,I pick eight models specifically to go through the process of clusering segmentation.
In this paper,I apply the K-means clustering algorithm to complete an approximate meaningful segmentation of several three-dimensional grid model.Initially,Floyd algorithm is employed to calculate arbitrary distance from any patch to any other patches of three-dimensional grid model under the environment of visual studio 2005.According to the result of the distances，I complete an approximate meaningful segmentation of several three-dimensional grid model,which is accomplished in the MATLAB platform.There is a comparison and detailed analysis between Dijkstra algorithm and Floyd algorithm.Realistic rendering of several three-dimensional grid model's segmentation results is realized in the MeshLab platform.Finally,specific explanation is listed in tables in order to show the results and rendering.
Keywords：
Three-dimensional grid model cluster segmentation Floyd algorithm realistic rendering
*若需了解更多与协助请咨询↓→[电脑QQ][手机QQ]【数据协助】
