多站点云自适应配准理论与技术研究

Point cloud registration is another key problem in the point cloud data processing technology of terrestrial laser scanning. With the theory of surveying adjustmeng, feature extraction and point cloud registration, the subject discusses the theory and method of the multiview cloud data adaptive registration by studying its own characteristics, which is based on the total least squares theory. Including: (1)Study the total least squares algorithm of targets' feature extraction and its accuracy evaluation model. Export the functional relationship between the accuracy of the feature extraction,scanning distance,scanning interval,spot diameter, incident angle, etc; (2)Study the total least squares algorithm of multiview cloud adaptive registration model. Export the consensus assessment method of correspondence lines feature and correspondence planes feature. At the same time, construct the adaptive registration model of multiview cloud containing some targets which is feature point, feature line or feature face, the adaptive registration model of multiview cloud containing no target, the adaptive registration model of multiview cloud which include some point cloud containing some targets and others containing no target; (3)Study the accuracy evaluation model of multiview cloud registratioin. Export the accuracy evaluation model of the whole point cloud, which have been registed by different adaptive registration algorithms. This topic will solve the bottleneck problem of theory and application in the multiview cloud adaptive registration, perfect the theory of the total least squares and the multiview cloud adaptive registration, promote the 3D laser scanning technology having a wide applications,and provide theoretical basis for making its measuring standard and developing its business software.

点云配准是地面三维激光点云数据处理技术中的关键问题之一，本课题从测量平差理论、特征提取理论和点云配准理论入手，通过研究点云数据自身的特性，在整体最小二乘理论基础下，探讨多站点云数据自适应配准技术的理论和方法。具体包括：（1）研究基于整体最小二乘的标靶特征提取算法及其精度评价模型，导出特征提取精度与扫描距离、扫描间隔、光斑大小、入射角等因素的函数关系；（2）研究基于整体最小二乘的多站点云自适应配准模型，导出同名线和面特征的一致性评价方法，建立多类特征标靶点云、多站无标靶点云、多站标靶和无标靶点云的自适应配准模型；（3）研究多站点云配准成果质量评价模型，推导多站点云自适应整体最小二乘配准的精度评价模型。 本课题的研究成果，将解决多站点云数据自适应配准理论与技术中的瓶颈问题，完善整体最小二乘理论和点云配准理论，促进三维激光扫描技术的广泛应用，并为制定相关测量规范和开发商用软件提供理论依据。

本项目研究内容涉及特征提取、标靶分布、配准模型、模型估计、精度评价、成果质量分析等六个方面，其主要成果如下：（1）提出了基于球体几何关系的球标靶特征提取算法，为球标靶特征提取研究提供了一种新的思路；（2）提出了标靶分布质量的定量评价模型，使分布质量的评价由定性升为定量，可指导用户布设标靶；（3）提出了一种基于“旋转矩阵误差阵的各元素平方和最小”的单闭合环多站无标靶点云整体配准算法，为无标靶点云的整体配准提供了一项新方案；（4）提出了标靶点云配准精度的严密评价模型，为所有扫描数据处理软件提供一种正确评价“点云处的配准精度”的方法；（5）提出了点特征信息量度量模型和点云信息量度量模型，给出了度量点云信息量的思路；（6）提出了点位误差椭球评价模型和点位误差平均椭球评价模型，给出了一种点云质量评价模型。.项目发表论文4篇（3篇SCI，1篇EI），录用论文1篇（核心），已投论文3篇（2篇英文，1篇中文），申请发明专利1项。