结构总体最小二乘估计三维坐标转换参数算法.docVIP

结构总体最小二乘估计三维坐标转换参数算法胡川1陈义1,2赵佩铭1于松松1公羽1张月超1祁欣欣1刘云鹏1 （1.同济大学测绘与地理信息学院上海 200092） 2.现代工程测量国家测绘局重点实验室上海200092） 摘 要 针对三维坐标转换参数估计过程中系数矩阵重复元素常规估计方法不能获得相同改正值的情况，提出采用结构总体最小二乘进行参数估计。把结构总体最小二乘问题看作是观测误差关于系数矩阵误差改正量和待估计参数的非线性最小二乘问题，线性化以后采用拉格朗日极值条件推导其迭代格式，并针对坐标转换的特殊情况设计系数误差改正矩阵的算法，最后模拟实验表明结构总体最小二乘法估计精度比总体最小二乘法稍低，比最小二乘法高，但是其能够保证系数矩阵中的重复元素获得相同改正值（零元素无改正值），理论比总体最小二乘法严密。 结构总体最小二乘；非线性最小二乘；三维坐标转换 An Iterative Algorithm for Structured Total Least Squares AdjustmentHU Chuan1, CHEN Yi1,2, ZHAO Pei-ming1,YU Song-song1, GONG Yu1,ZHANG Yue-chao1,QI Xin-xin1,LIU Yun-peng1 （1. College of Surveying and Geo-informatics, Tongji University, Shanghai 200092） （2. Key Laboratory of Modern Engineering Surveying, SBSM, Shanghai 200092） Abstract：The structured total least squares not only considers the perturbation of coefficient matrix A in equation but also keep the structure of error matrix agree with it, let the adjustment theory more strictly. In this paper, it considers the structured total least squares as traditional nonlinear least squares according to the conventional expression of the condition constrained adjustment method, and regards the errors matrix as party of the parameters to be estimated. The Lagrange multiplier is exploited to derive the iteration function of structured total least squares since the nonlinear function that related with and is deployed by Taylor series. A numerical example is given at last. Key words：Geodesy and Survey Engineering；Total Least Squares; Structured Total Least Squares; Nonlinear Least Squares;3D- Coordinates Translation 1 引言 总体最小二乘法（Total Least Squares TLS）是在最小二乘法顾及系数矩阵误差情况下的一种自然扩展。TLS这一术语最先出现在由G. H. Golub和C. F. Van Loan 两人于1980年发表在《SIAM Journal on Numerical Analysis》上的An Analysis of The Total Least Squares Problem”一文中[1]。经过多年的发展，总体最小二乘在工程实践领域得到了广泛的应用，例如图像恢复、模式识别、信号处理等。在实践中发现，有时系数矩阵元素具有重复性，同一元素可能多次出现在矩阵的不同位置。1993年De Moor[2]将这种具有结构特性的总体最小二乘方法称作结构总体最小二乘法（Structured Total Least Squares STLS）。结构总体最小二乘问题不仅出现在上述的工程实践领域中[3-8]，而且在测绘工程领域也时常出现，如坐标转换参数估计[