近景摄影测量中的光束法平差快速算法

由于近景摄影测量的摄影方式灵活多变,光束法平差中构成的法方程系数矩阵不具有带状结构,导致基于整体广义逆法的求解速度很慢。为解决这一问题,提出了一种新的光束法平差快速解法。该方法以"R-T"模式的共线方程为基础,首先构建摄像机参数的法方程式,以整体解算摄像机参数的改正数,然后再逐点解算三维点改正数。采用该方法避免了组建整体误差方程式和三维点法方程式,能在保证精度的前提下有效提高计算速度。仿真实验和真实图像实验均表明,本文算法具有较强的可行性和鲁棒性,特别适合拍摄角度多变的近景摄影测量系统。

Fast Algorithm for Bundle Adjustment Based on Close-range Photogrammetry

As the shoot methods are flexible in close-range photogrammetry,the banded structure is not included in the coefficient matrix of normal equation that formed with bundle adjustment.Thus,computing speed based on generalized inverse method is slow.To solve the problem,a new fast algorithm based on the collinearity equation of R-T-mode was proposed.Firstly,the normal equation of camera parameters was built,followed by calculating its corrections.Secondly,the 3D corrections were calculated by point-to-point,thus the formations of global error equation and normal equation of 3D point can be avoided,meanwhile,computing speed can be improved efficiently without loss precision.Through both simulation experiment and real image experiment,it was testified that the algorithm has a superior feasibility and robustness,particularly suitable for the close-range photogrammetry system with angle-varied shoots.