摘要
为了实现对特大钢结构的高效率、高精度检测,对地面激光点云整体配准算法以及无人机多视影像生成密集点云算法进行了研究。采用基于几何特征的可迭代整体配准算法不断对观测值进行约束定权和解算,将观测值改正数误差控制在一定阈值范围内,直到完成配准,生成整个网架结构的激光雷达点云模型。以球节点多连杆中心点算法对网架结构中的球结点与柱进行偏心计算,再使用基于视觉运动恢复结构算法以及改进的RANSAC算法生成的影像密集点云,通过配准实现地面激光点云和高分辨率非量测影像数据的融合。以某亚洲最大跨钢结构高精度检测为例,多站激光雷达点云整体的配准精度为5 mm,抽检的21根钢柱里16根柱子的偏差接近或大于35 mm(32.1~68.2 mm,方向均向外),整体钢结构的网架挠度均小于1/250。由此表明,地面激光点云整体配准算法以及高分辨率非量测影像数据密集点云生成算法可行且准确,能够满足特大钢结构的高精度检测要求。
To achieve high-efficiency and high-precision detection of super-large steel structures,this study investigates the overall registration algorithm for ground laser point clouds and the algorithm to generate dense point clouds from unmanned aerial vehicle multi-view images.First,the iterative global registration algorithm based on geometric features is used to weigh and solve the observed value constraints continuously.The error of the observed value corrections is controlled within a certain threshold range until the registration is complete and the LiDAR point cloud model of the entire grid structure is generated.Subsequently,the eccentricity between the spherical node and the column in the grid structure is calculated using the spherical node multi-link center point algorithm.The image dense point cloud generated using the visual structure from motion algorithm and improved RANSAC algorithm is used to realize fusion of the ground laser point cloud and high-resolution non-metric image data through registration.With the high-precision inspection of the largest-span steel structure in Asia taken as an example,the overall registration accuracy of the multi-station LiDAR point cloud is 5 mm,and the deviation of 16 of the 21 steel columns sampled is close to or greater than 35 mm(32.1-68.2 mm,all in the outward direction).Notably,the deflection of the overall steel structure grid is less than 1/250.The feasibility and accuracy of the entire registration algorithm for the ground LiDAR point cloud and the algorithm for generating high-resolution nonmeasurement image data-intensive point clouds are verified to fully meet the requirements of high-precision detection of super-large steel structures.
作者
郭明
孙梦溪
黄明
闫冰男
周玉泉
赵有山
GUO Ming;SUN Meng-xi;HUANG Ming;YAN Bing-nan;ZHOU Yu-quan;ZHAO You-shan(School of Geomatics and Urban Spatial Informatics,Beijing University of Civil Engineering and Architecture,Beijing 100044,China;Representative Architecture and Ancient Architecture Database Engineering Research Center of Ministry of Education,Beijing 102616,China;Key Laboratory of Urban Spatial Information,Ministry of Natural Resources,Beijing 102616,China;China Academy of Building Research,Beijing 100013,China)
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2021年第5期989-998,共10页
Optics and Precision Engineering
基金
国家自然科学基金资助项目(No.41971350)
北京建筑大学金字塔人才培养工程建大杰青培养计划资助项目(No.JDJQ20200307)
北京市未来城市设计高精尖创新中心资助项目(No.UDC2019031724)
研究生创新项目(No.PG2020082,No.PG2020075)。
关键词
近景摄影测量
激光雷达
无人机
钢结构检测
点云
close-range photogrammetry
lidar
unmanned aerial vehicle
steel structure detection
point cloud
