1.College of Civil Engineering,Xi’an University of Architecture and Technology,Xi’an 710055,China
2.Key Laboratory of Structure Engineering and Earthquake Resistance,Xi’an University of Architecture and Technology,Xi’an 710055,China
3.Industrial Construction Division,Inspection and Certification Co. ,Ltd. ,MCC,Beijing 100088,China
4.College of Information and Control Engineering,Xi’an University of Architecture and Technology,Xi’an 710055,China
Show less
文章历史+
Received
Published
2025-03-26
2026-01-25
Issue Date
2026-02-12
PDF (3114K)
摘要
为分析集成三维点云逆向建模方法(integrated 3D point cloud reverse modeling,IPCRM)在建立薄厚型钢构件三维模型时的精度表现,以局部变形角钢为研究对象,利用SfM (structure from motion)-MVS (multi-view stereo)算法建立其三维点云模型,借助逆向建模技术生成曲面模型,重点开展了模型精度验证试验. 结果表明:各表面形状特征参数的相对误差均在8%以内(吻合度验证);4种角钢模型与实际角钢间无显著性差异(P值,P=0.99),且角钢厚度对模型精度无显著性影响(P值,P=0.95),结论在95%的置信度水平下成立(差异显著性验证). 研究结果为后续算法优化及利用此类方法进行合理的钢构件局部变形损伤检测与承载性能评价提供依据.
Abstract
To analyze the precision of the integrated 3D point cloud reverse modeling (IPCRM) method in generating three-dimensional models of thin-thickness steel members, locally deformed angle steels are taken as the research objects. Three-dimensional point cloud models of locally deformed angle steels are established using the SfM (structure from motion)-MVS (multi-view stereo) algorithm, and the surface models of locally deformed angle steels are generated with the help of reverse engineering technology. Special attention is placed on model precision verification experiments. The results show that: the relative errors of all surface shape characteristic parameters are within 8% confirming the conformity of the models; there is no significant difference between the four angle steel models and the actual angle steels (P value, P=0.99), and angle steel thickness has no significant influence on model precision (P value, P=0.95). These conclusions hold at a 95% confidence level (significance verification). The research results provide a basis for subsequent algorithm optimization and the rational use of this method for local deformation damage detection and bearing performance evaluation of steel members.
三维重建可视为逆向投影过程:通过光心与像素点构建投影射线,多视图中同名像点对应的射线在理想状态下应汇聚于同一空间点. 在已知相机内参数的条件下,光心至像素点的投影映射可线性表达为确定的空间射线方程. 为实现多射线交会约束,需将各影像统一至同一坐标系,其核心问题在于求解视图间的相对位姿参数. 这一几何关系在对极几何框架中由基础矩阵 F 与本质矩阵 E 表征,其中 E 适用于已知内参数的理想情况, F 则适用于更普遍的未标定情形.
式中: K 是相机的内参矩阵(或定标矩阵),包含了fx 、fy 、s、u0和v0等参数,fx 和fy 分别为图像的横向和纵向缩放因子,s为非矩形像素引起的倾斜因子,(u0,v0)T为相机光轴与图像平面的交点的像素坐标;[ RWC | t ]3×4表示点M所处的世界坐标系转化为相机坐标系所需的旋转和平移矩阵.
2) 点特征检测与匹配.SIFT算法[29]使用高斯核函数对原图像进行上、下采样,构造高斯金字塔尺度空间.使用差分高斯(difference of Gaussian, DoG)捕捉图像中目标轮廓,近似计算高斯-拉普拉斯算子(Laplacian of Gaussian, LoG),以提高计算效率. SIFT点特征检测方法是图像的局部特征检测算法,其对旋转、尺度缩放、亮度变化保持不变性,对视角变化、放射变化及噪声也保持一定程度的稳定性. 点特征检测后,SIFT为每个关键点生成包含位置、尺度方向等信息的描述符,通过特征匹配算法(如暴力匹配、快速最近邻等)建立多视影像间的同名点对应关系,匹配点需满足对极几何约束条件.
3) 对极几何、基础矩阵和本质矩阵.对极几何约束描述的是两幅视图之间的内在射影关系,与外部场景无关,只依赖摄像机内参数和两幅视图之间的相对位姿. 通过施加对极几何约束,可以极大程度减少特征匹配的搜索范围. 如图2所示,同名点m1和m2的约束关系由基础矩阵 F 描述[式(3)]. 基于 F 矩阵,匹配时只需沿极线在一定阈值内搜索同名点(实际受噪声影响,对应点不会严格位于极线上).
本质矩阵 E 与基础矩阵 F 的核心区别在于坐标系与作用域: E 在相机坐标系下描述视图间的旋转、平移关系; F 在像素坐标系下关联两图像像点,并考虑相机内参数的影响,实质是将本质矩阵从相机坐标系转换到像素坐标系下得到的矩阵,式(4)~式(6)描述了这一过程.
对于非定标视图基本矩阵的估计,由式(3)可以得到:
若已知8个对应图像点对,联立8个线性方程可获得如下线性方程组:
为求解上述方程,建立线性方程组 Af =0,由系数矩阵 A 的最小奇异值对应的奇异向量 f 求出初始线性解 F . 根据奇异性约束,使二范数取得最小值的作为最终解. 之后,通过式(5)反推本质矩阵. 此外,由于实际情况中可能存在一定的误差,导致同名点可能没有精准地落在极线上,而是落在一个阈值区间内,需要通过RANSAC算法进行优化,缩小误差范围. 最后,通过求解图像之间的相对位姿,可重建三维稀疏点云模型.
ZHENGT X, HUANGS, LIY F,et al .Key techniques for vision based 3D reconstruction:a review[J].Acta Automatica Sinica,2020,46(4):631-652.(in Chinese)
[3]
ISGROF, ODONEF, VERRIA .An open system for 3D data acquisition from multiple sensor[C]//Seventh International Workshop on Computer Architecture for Machine Perception (CAMP’05). July 4-6,2005,Palermo,Italy.IEEE,2005:52-57.
[4]
董秀军 .三维激光扫描技术及其工程应用研究[D].成都:成都理工大学,2007.
[5]
DONGX J .The three-dimensional laser scanning technique and research on its engineering application[D].Chengdu:Chengdu University of Technology,2007.(in Chinese)
[6]
张宝誉 .基于结构光的复杂场景深度获取[D].西安:西安电子科技大学,2018.
[7]
ZHANGB Y .Study on structured light based depth sensing for complex scenes[D].Xi’an:Xidian University,2018.(in Chinese)
[8]
PARKJ, KIMH, TAIY W,et al .High quality depth map up sampling for 3D-TOF cameras[C]//2011 International Conference on Computer Vision.November 6-13,2011,Barcelona,Spain.IEEE,2011:1623-1630.
LIQ Q, LIB J, CHENJ .Research on laser range scanning and its application[J].Journal of Wuhan Technical University of Surveying and Mapping, 2000, 25(5): 387-392.(in Chinese)
[11]
魏祖帅 .倾斜摄影空中三角测量若干关键技术研究[D].焦作:河南理工大学,2015.
[12]
WEIZ S. The key technology of aerial triangulation for oblique photogrammetry[D]. Jiaozuo:Henan Polytechnic University,2015.(in Chinese)
MIAOT, XUT Y, DENGH B,et al. Development of an appearance texture generation system for single plant organ image[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE),2020,36(7):208-216.(in Chinese)
YANGX B, HUANGX B .Surface shape reconstruction of wrinkle fabric based on shape from shading[J].Journal of Image and Graphics,2003,8(2): 181-187.(in Chinese)
[17]
王刚 .基于多视角立体视觉的三维重建研究[D].哈尔滨:哈尔滨工业大学,2019.
[18]
WANGG. Research on 3D reconstruction based on multi-view stereo[D]. Harbin:Harbin Institute of Technology,2019.(in Chinese)
HUP C, GUOY, LIB G,et al .Three-dimensional reconstruction and its precision evaluation of plant architecture based on multiple view stereo method[J]. Transactions of the Chinese Society of Agricultural Engineering,2015,31(11): 209-214.(in Chinese)
LIUY F, FANJ S, KONGS Y,et al .Detection of structural defect and deformation based on multi-view geometric three-dimensional reconstruction method[J].Engineering Mechanics,2020,37(9):103-111.(in Chinese)
[23]
张茂正 .面向高耸古塔的数字化三维实景建模方法研究[D].银川:宁夏大学,2023.
[24]
ZHANGM Z .Research on digital 3D real scene modeling method for towering ancient pagoda[D].Yinchuan:Ningxia University,2023.(in Chinese)
[25]
XUY, ZHANGJ .UAV-based bridge geometric shape measurement using automatic bridge component detection and distributed multi-view reconstruction[J].Automation in Construction,2022,140:104376.
[26]
ZHAOS Z, KANGF, LIJ J .Concrete dam damage detection and localisation based on YOLOv5s-HSC and photogrammetric 3D reconstruction[J]. Automation in Construction,2022,143:104555.
[27]
GAOW H, ZHANGC Y, LUX L,et al. Concrete spalling damage detection and seismic performance evaluation for RC shear walls via 3D reconstruction technique and numerical model updating[J]. Automation in Construction, 2023, 156: 105146.
ZHOUY, MENGS Q, KONGQ Z,et al .State-of-the-art of intelligent damage detection and response prediction of building structures[J].Journal of Building Structures,2024,45(6):107-132.(in Chinese)
ZHOUY, LIUP, HAOG W,et al .Method on inverse establishment of bridge finite element model based on UAV oblique photography technology[J].Journal of Hunan University (Natural Sciences),2023,50(9):13-23.(in Chinese)
[32]
ZHENGJ H, QIANK, ZHANGD T .Reverse reconstruction of geometry modeling and numerical verification of 2.5D woven composites based on deep learning[J].Composite Structures,2024,329:117801.
[33]
ROSHANT R, JAFARIM, GOLAMIM,et al .Evaluating geometric measurement accuracy based on 3D model reconstruction of nursery tomato plants by Agisoft photoscan software[J].Computers and Electronics in Agriculture,2024,221:109000.
[34]
SCHONBERGERJ L, FRAHMJ M .Structure-from-motion revisited[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).June 27-30,2016.Las Vegas,NV,USA.IEEE,2016:4104-4113.
[35]
HUANGT S, NETRAVALIA N. Motion and structure from feature correspondences:a review[J]. Proceedings of the IEEE,82(2):252-268.
HEG X, MAJ W, ZHANGX K,et al .An improved image mosaic algorithm based on SURF and RANSAC[J]. Applied Science and Technology,2018,45(4):45-49.(in Chinese)
[38]
康来 .基于图像点特征的多视图三维重建[M].北京:科学出版社,2015.
[39]
KANGL .Multi-view 3D reconstruction based on image feature points[M].Beijing:Science Press,2015.(in Chinese)
[40]
YANGJ L, LIH D, JIAY D .Optimal essential matrix estimation via inlier-set maximization[M]//Computer Vision-ECCV 2014.Cham:Springer International Publishing,2014: 111-126.
[41]
FAUGERASO D, LUSTMANF .Motion and structure from motion in a piecewise planar environment[J]. International Journal of Pattern Recognition and Artificial Intelligence, 1988, 2(3): 485-508.
JIAOC, WANGB F, YIY H. Research on point cloud data filtering algorithms[J]. Foreign Electronic Measurement Technology,2019,38(11):18-22.(in Chinese)
[49]
RUSUR B, MARTONZ C, BLODOWN,et al .Towards 3D Point cloud based object maps for household environments[J]. Robotics and Autonomous Systems, 2008, 56(11): 927-941.
HAND S, XUM L, JINY H .Filtering and accuracy analysis of multi-source heterogeneous point cloud registration data[J].Journal of Geomatics Science and Technology,2020,37(5):503-508.(in Chinese)