无人机遥感影像大规模光束法平差快速解算

Fast Calculation of Large-scale Bundle Block Adjustment Based on UAV Remote Sensing Image

为实现对无人机遥感影像大规模光束法区域网平差的快速解算,采用逐点消元法并利用分块稀疏矩阵保存消元后的法方程,以减少内存的使用量;利用预处理共轭梯度算法实现快速解算,利用OpenMP技术实现预处理共轭梯度迭代计算多核并行处理。结果表明,将消元法和预处理共轭梯度算法用于无人机遥感影像大规模光束法区域网平差解算,既可节省内存,又可提高计算效率。

In order to achieve the fast calculation of large-scale bundle block adjustment based on UAV remote sensing image, a method of point wise elimination was presented in this paper to reduce storage usage. The normal equations obtained after elimination were stored by block sparse matrix to further reduce storage usage. Pre-conditional conjugate gradient(PCG) was used to achieve fast solution.OpenMP technology was used for multi-core synchronization processing of PCG.The experimental result shows that using elimination method and PCG to calculate the large-scale bundle block adjustment based on UAV remote sensing image, can save storage and improve computational efficiency.