光束法平差是一种高精度的摄影测量优化方法, 在大尺寸三维测量中起不可替代的重要作用。但光束法平差优化参数初始值的误差会影响最终结果的优化精度、优化发散以及局部收敛等现象发生。以高精度的区域网光束法平差为研究对象, 系统性的对包括重建三维点坐标、图像坐标、外参数旋转角和平移向量四种初值参数对测量精度的影响进行深入分析。模拟研究表明, 光束法平差优化将毫米级的重建三维点初值误差降低10倍以上; 随着图像坐标定位误差的增加优化精度下降很快, 但光束法平差优化精度仍能提高一个数量级; 相机外参数旋转角和平移向量初值误差都对光束法平差优化精度影响较大。模拟结论对实际测量实验具有重要的指导意义。

Bundle adjustment is a high-precision photogrammetry method of optimization, and plays an irreplaceable role in large-scale 3D measurement. However, the error of the initial value of the bundle adjustment parameters will cause the bad accuracy of the result, optimized divergence and local convergence. The influence of four initial parameters including reconstruction of three-dimensional point coordinates, image coordinates, external parameter rotation angle and translation vector on the measurement accuracy is analyzed, which based on the high-precision bundle adjustment of regional network. Simulation results show that the bundle adjustment optimization improves the initial accuracy of the reconstructed 3D point of the millimeter level by more than 10 times. With the increase of the image coordinate error, the optimization precision decreases rapidly, but the accuracy of bundle adjustment still can be increased by an order of magnitude. Both of the external rotation angle and the translation vector have a great influence to the optimization accuracy of the bundle adjustment. The simulation conclusion have important guiding significance for the actual measurement.