组特征拓扑关系约束的倾斜影像直线可靠匹配方法

Three-dimensional building reconstruction by using oblique images has encountered severe problems in maintaining accurate building edge information. Straight line correspondence is a kind of good feature to constrain the reconstruction and generate fine three-dimensional model. However, it is difficult to obtain reliable straight line matching results for oblique images using existing methods due to the problems in consistent corresponding feature support regions computation，parallel lines recognition and broken lines extraction. Aiming at improving the straight line matching performance for oblique images, lines with constraints are grouped and regarded as matching primitives in this project. The main researches of this project include: 1) Designing line-pairs producing and recording method which considers building structure and matching redundancy, and improving the matching reliability based on matching redundancy produced by a line assigned to different line-pairs; 2) Exploring epipolar and topological constraints from the intersection point of the two line segments in line pairs and ensuring the descriptor construction of consistent support regions under non-strict coplanar condition constraint to achieve the reliable matching of line-pairs; 3) Proposing a matching result recording method based on a double relation matrix to integrate multi-source information. Then, redundancy matching and linear topology are combined to check the matching results for outlier elimination; 4) Developing prototype system and perform typical comparative experimental analysis. This project is devoted to dramatically improve the straight line matching of oblique images to promote the development and the application of oblique photogrammetry in 3D urban modeling.

面向倾斜影像建筑物边缘的精细三维建模，突破现有城区倾斜影像直线匹配过程中建筑物边缘同名直线特征区域不一致、建筑物立面细节处邻近平行直线局部特征重复使用、直线提取断裂引起的错误匹配及匹配结果难以检核的瓶颈。本项目采用编组的直线对作为匹配基元，主要研究内容包括：1)顾及建筑物结构特征和可产生匹配冗余的直线对提取及记录方式，利用同一直线与其他不同直线编组产生的匹配冗余提高匹配的可靠性；2)挖掘直线对内两直线交点为主的核线约束及拓扑约束，确保非严格共面条件约束下一致性支撑区的描述子构建，完成直线对的可靠匹配；3)提出一种集成多源信息的双层关系矩阵的匹配结果记录方式，在此基础上，综合匹配冗余和直线拓扑实现匹配结果的有效检核，确保高可靠的匹配结果；4)原型系统设计研发与典型实验分析。本项目研究将为倾斜影像直线匹配提供有效的理论依据和关键技术支撑，促进倾斜影像在三维城市建模中的发展和应用。

本项目针对倾斜影像直线匹配过程中建筑物边缘难以有效匹配、直线匹配结果复杂难以检核的问题，研究从直线匹配基元入手，提取顾及建筑物结构特征的直线对用于匹配，挖掘线对内两直线间有效几何信息用于约束线对匹配，寻求易于记录及结果整合的线对匹配结果记录方式和有效的匹配结果检核算法。研究突破倾斜影像直线匹配难点问题，为基于影像的三维模型重建提供可靠的线特征支持。主要工作包括以下几个关键部分：.(1) 线对提取。在满足邻近直线及角度约束的基础上，采用可变的自适应搜索区域替代已有研究中的固定搜索区域，确保每条直线都参与编组及匹配。.(2) 线对匹配。在获取线对中两直线交点核线的基础上，依次采用核线约束、单应矩阵约束、局部方位约束确定匹配候选。在此基础上构建基于线对的Daisy描述子用于确定最优匹配对，得到参考影像、搜索影像上的同名直线组及同名直线。 .(3) 匹配结果检核。在利用双层关系矩阵记录匹配结果的基础上，提出一种结合匹配冗余和共线约束的匹配结果检核算法，有效地解决了“一对多”、“多对一”、“多对多”结果中的匹配冲突问题、保留正确匹配，剔除错误匹配，最终实现组直线匹配结果的有效检核及整合。.(4) 直线匹配应用。为了确保准稠密匹配结果点云中建筑物边缘的规则性，研究在已有同名点构建三角网的基础上，利用匹配直线对三角网进行优化，提出一种基于优化三角网约束的准稠密匹配算法，以三角形内切圆圆心为初始加密点，实现匹配传播及迭代加密匹配。通过对优化前后三角网匹配结果进行对比分析，验证利用直线约束准稠密匹配的有效性。

内容获取失败，请点击重试