摘要
针对基于传统平差方法的建筑物边界规则化建模和解算复杂问题,利用基于海森法线和正交距离的直线方程,构建顾及边界垂直和平行条件的变量含误差(EIV)模型.该模型是附有二次型限制条件的齐次EIV模型,且设计矩阵由含误差元素的矩阵和不含误差的矩阵组成.针对该模型,推导了基于奇异值(SVD)分解的混合最小二乘-总体最小二乘(LS-TLS)直接解法以及精度评定方法.通过LiDAR点云的建筑物边界规则化算例表明:该方法易于建模,避免了条件方程线性化问题,且解算不需要迭代,重建精度高,可广泛用于基于摄影测量和LiDAR数据的建筑物重建.
Considering the modeling and computational complexity of building boundary regularization task of the traditional least squares adjustment methods,an EIV(errors-in-variables)model considering the vertical and parallel relations of building boundary was presented by using Hessian normal vector and orthogonal distance's line equation.The proposed model is a homogeneous EIV model with a quadratic condition,and the design matrix consists of a matrix with random errors and an error free matrix.A direct adjustment method based on singular value decomposition(SVD)of mixed least squares-total least squares(LS-TLS)with accuracy evaluation was proposed.The building boundary regularization from LiDAR point cloud shows that the proposed modelling and numerical algorithm can avoid solving a complicated nonlinear conditional equations and have the advantages of simplicity of modeling,stability and efficiency in numerical computation,and better accuracy.It can be widely used for building reconstruction from photogrammetry or LiDAR data points.
作者
周瑜
胡莘
周拥军
王建荣
ZHOU Yu;HU Xin;ZHOU Yongjun;WANG Jianrong(Institute of Surveying and Mapping,Information Engineering University,Zhengzhou,Henan 450001,China;State Key Laboratory of Geo-Information Engineering,Xi'an Research Institute of Surveying and Mapping,Xi'an,Shaanxi 710054,China;Schoo1of Naval Architecture,Ocean and Civil Engineering,Shanghai Jiao Tong University,Shanghai 200240,China)
出处
《中国矿业大学学报》
EI
CAS
CSCD
北大核心
2019年第2期452-458,共7页
Journal of China University of Mining & Technology
基金
国家自然科学基金项目(41274012)
工信部民机专项基金项目(MJZ-2016-S-44)
