基于区域分割的三维点云模型骨架提取算法研究

发布时间：2019-02-23 20:43

【摘要】：骨架是概括表示三维模型的一种抽象化手段,它能有效刻画并表达模型的拓扑结构和形状特征,同时节省模型在计算机上的存储空间,因此骨架提取技术被广泛应用于各领域。但是传统算法仍存在骨架提取结果易产生错位或忽略模型细节特征的问题。本文在计算机辅助文物保护与复原的研究课题基础上,结合三维点云模型的分割技术,针对点云模型的骨架提取问题展开了一系列的研究,主要工作如下:(1)提出一种基于马尔科夫随机场(Markov Random Filed,MRF)的点云模型分割算法。针对传统分割算法容易产生过分割的问题,该方法将马尔科夫随机场模型应用于点云模型分割,利用点云模型几何特征作为先验信息,根据点云属性和邻域点相关性确定目标函数并求解得到点的最优标号集,完成点云模型的有效分割。实验结果表明,该算法在保留点云模型特征的基础上,提高了分割过程的自适应性和分割结果的准确性。(2)提出一种基于区域分割的点云模型骨架提取算法。针对传统方法在相邻区域较紧密模型中提取骨架结果易跨越区域的问题,该方法在分割后区域相邻处生成连接点,同时在区域内自适应计算初始收敛邻域尺度并提取各区域骨架分支,最后通过连接点将相邻区域骨架分支连接得到完整的骨架。实验结果表明,该方法减少了全局迭代收敛的计算量和迭代次数,同时使用局部稠密度对点云密度加权,降低噪声点的影响程度,保证其拓扑结构正确。除此之外,通过引入连接角的概念并建立新的连接方式,确保提取骨架的完整性。(3)本文算法主要应用于“基于全局优化的破损兵马俑虚拟复原方法研究”项目,针对秦俑点云模型进行骨架提取,得到符合原始模型特征的骨架,以此支撑后续模型的拓扑结构分析以及秦俑文物的形状匹配与识别。
[Abstract]:Skeleton is an abstract method to represent 3D model. It can effectively describe and express the topological structure and shape features of the model and save the storage space of the model on the computer. Therefore skeleton extraction technology is widely used in various fields. However, the traditional algorithm still has the problem that the skeleton extraction results are prone to misposition or ignore the detailed features of the model. Based on the research of computer aided cultural relic conservation and restoration and the segmentation technology of 3D point cloud model, a series of research on the skeleton extraction of point cloud model is carried out in this paper. The main work is as follows: (1) A point cloud model segmentation algorithm based on Markov Random Field (Markov Random Filed,MRF) is proposed. In view of the problem that the traditional segmentation algorithm is easy to produce over-segmentation, the Markov random field model is applied to the point cloud model segmentation, and the geometric feature of the point cloud model is used as the prior information. The objective function is determined according to the attribute of point cloud and the correlation of adjacent points, and the optimal label set of points is solved to complete the efficient segmentation of point cloud model. The experimental results show that the algorithm improves the self-adaptability of the segmentation process and the accuracy of the segmentation results on the basis of preserving the features of the point cloud model. (2) an algorithm for extracting the skeleton of the point cloud model based on regional segmentation is proposed. In order to solve the problem that the traditional method can easily cross the region by extracting the skeleton in the close-knit model of adjacent regions, the proposed method generates joint points at the adjacent regions after segmentation. At the same time, the initial convergence neighborhood scale is calculated adaptively in the region, and the skeleton branches of each region are extracted. Finally, the complete skeleton is obtained by connecting the skeleton branches of the adjacent regions. Experimental results show that the proposed method reduces the computational complexity and iteration times of global iteration convergence, and uses local density to weight point cloud density, reduces the influence of noise points, and ensures the correct topology structure. In addition, by introducing the concept of connection angle and establishing a new connection way to ensure the integrity of skeleton extraction. (3) this algorithm is mainly used in the project of "virtual restoration method of damaged Terracotta Warriors and horses based on global optimization". The skeleton is extracted from the point cloud model of the terracotta warriors, and the skeleton is obtained according to the characteristics of the original model, so as to support the topological structure analysis of the subsequent model and the shape matching and recognition of the cultural relics of the terracotta warriors.
【学位授予单位】：西北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】