A planar minimax algorithm for analysis of coordinate measurements (Q1895919): Difference between revisions

Two sets of \(N\) points in the plane are given -- a fixed set of nominal points \(z_1,\dots, z_n\) and a set of measured points \(\widehat x_1,\dots, \widehat x_n\). The following problem of nonlinear minimax approximation is considered: transform the measured points as a whole, by translation and rotation so that the maximal Euclidean distance between corresponding points in the two sets is minimized, that means \(\max|x_k- z_k|= \min!\), where \(x_k= A_0\widehat x_k+ u\) with \(k= 1,\dots, n\) and suitable \(A_0\) and \(u\). A solution of a least-squares version of this problem is known. Here, in chapter 2 the problem is analyzed and some theorems characterizing the solution are proved. An algorithm to solve the problem under consideration is developed here. It is tested using MATLAB and C for an AT \& T 386 SX PC. Then the case in which the tolerances are allowed to vary from point to point is studied. It is indicated how the algorithm should be modified to handle weights. Optimal joint scaling is dealt with. A special approximation problem is observed from the standpoint of linear algebra.