基于多时相点云的贴金石质文物病害检测

发布时间：2018-05-11 16:03

本文选题：多时相点云 + 金箔病害　；参考：《北京建筑大学》2013年硕士论文

【摘要】：对于贴金石质文物的修复工作，最为重要的是对金箔的修复，这种工作难度大，工序复杂，费时较长，且耗资巨大。例如重庆大足宝顶山贴金千手观音造像的修复，单单金箔修复的工序就有二十几道，历时六年，耗资千万。因此对于贴金石质文物的保护，一个非常重要的方面就是对金箔的有效保护，而其前提是对金箔病害进行有效的病害检测。本文正是以此为切入点，利用三维激光扫描技术，通过采集多时相的点云数据，进行多时相点云的对比检测，从而发现金箔的病害并量算病害面积，，为后续文物修复与保护工作提供数据支撑。贴金石质文物病害提取的一个关键就是确定病害的发生阈值。本文在分析病害阈值组成的基础上，导出了病害阈值的公式，并从仪器误差、数据采集误差与数据处理误差三个方面进行了系统的分析，从而指出了提高点云精度应采取的具体措施。贴金石质文物病害检测采用部分Hausdorff距离做为核心算法，基于病害发生阈值，对病害区域进行提取与标示。由于部分Hausdorff距离需要对整个点集进行运算，比较耗时。而线性八叉树是点云数据处理中一种较为常用的数据组织方式，该方式层次清楚，编码简单，占用内存小，若利用线性八叉树对点云数据进行组织，则对其邻域点云进行检索、运算即可。因此，本文提出了利用线性八叉树对点云进行组织的方法。在线性八叉树的基础上对编码方式进行了改进，将八进制的编码方式转换为自然数进行索引，提高了检索的效率。针对剖分层次的限制，提出了改进的剖分方式，提高了内存的利用率。在文物实际保护工作中，只给出病害的区域还不够。因此，本文提出了利用病害面积对病害区域进行量算的方法。利用三角格网模型面积对实物表面进行逼近，提出基于病害区域三角格网的搜索算法，利用海伦公式计算病害区域的面积。本文利用开发的贴金石质文物病害检测试验原型系统，以重庆大足宝顶山千手观音造像为试验对象，对提出的算法流程进行了验证，证明了本算法流程在贴金石质文物病害检测领域的可行性。
[Abstract]:The restoration of gold foil is the most important work for the restoration of gold stony relics. This kind of work is difficult, complex, time-consuming and costly. For example, there are more than 20 gold foil restoration procedures for the restoration of the gold thousand hands Guanyin statue on the Dazu Baoding Mountain in Chongqing, which lasted for six years and cost tens of millions of dollars. Therefore, an important aspect of the protection of gold foil is the effective protection of gold foil, which is based on the effective disease detection of gold foil. In this paper, by using 3D laser scanning technology and collecting multi-temporal point cloud data, we can find out the damage of gold foil and calculate the area of the disease. To provide data support for the subsequent restoration and protection of cultural relics. A key to the extraction of the disease is to determine the threshold of the disease. On the basis of analyzing the composition of disease threshold, the formula of disease threshold is derived, and the error of instrument, data acquisition and data processing are analyzed systematically. The concrete measures to improve the accuracy of point cloud are pointed out. Partial Hausdorff distance is used as the core algorithm to detect the disease of gold paste stone relics. Based on the threshold of disease occurrence, the disease area is extracted and marked. Because part of the Hausdorff distance needs to operate on the whole set of points, it is time consuming. The linear octree is a common data organization method in point cloud data processing. This method has the advantages of clear level, simple coding and small memory. If the linear octree is used to organize the point cloud data, the neighborhood point cloud can be retrieved. Operation is sufficient. Therefore, a method of organizing point clouds using linear octree is proposed. On the basis of linear octree, the coding method is improved, the octal coding method is converted to natural number for index, and the efficiency of retrieval is improved. Aiming at the limitation of partition level, an improved partition method is proposed to improve the utilization of memory. In the actual protection of cultural relics, it is not enough to give only the area of disease. Therefore, this paper puts forward a method to calculate the disease area by using disease area. The triangular grid model area is used to approximate the real surface, and a search algorithm based on the triangular grid of the disease region is proposed. The area of the disease area is calculated by using Helen's formula. In this paper, a prototype system for disease detection of gold-coated stone relics is developed, and the proposed algorithm flow is verified by taking the Qianhand Guanyin image of Dazu Baoding Mountain in Chongqing as the experimental object. It is proved that this algorithm is feasible in the field of disease detection of gold-coated stone relics.
【学位授予单位】：北京建筑大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：G264.3;P225.2

【参考文献】