稳健李代数旋转平均用于GPS辅助无人机影像三维重建

发布时间：2018-06-02 07:19

本文选题：无人机 + 三维重建　；参考：《光学精密工程》2017年06期

【摘要】：针对最小二乘的旋转平均方法对粗差敏感,求解影像旋转参数不够精确的问题,提出了一种稳健的旋转平均方法。先利用李群和李代数之间的映射关系,将旋转矩阵的乘积运算简化为李代数中的减法运算,推导出旋转平均迭代解算的线性化方程;然后利用L1范数优化和迭代加权最小二乘相结合的方法求解全局一致旋转最优解;最后采用迭代策略剔除粗差,得到精确的旋转矩阵。实验结果表明,与传统最小二乘方法相比,提出方法的旋转参数求解精度更高,稳健性更好,用于三维重建可以得到更密集均匀的点云,重建完整性更好。旋转平均的精度优于0.15度,计算时间不超过0.31s,光束法平差后,重投影误差在1.3个像素以内。基本满足快速稳健三维重建的要求。
[Abstract]:In order to solve the problem that the rotation average method of least squares is sensitive to the rough difference and solves the problem that the rotation parameters of the image are not accurate enough, a robust rotating averaging method is proposed. First, the product operation of the rotation matrix is simplified to a subtraction operation in the lie algebra by using the mapping relation between Li Qun and Lie algebra. By using the method of L1 norm optimization and iterative weighted least squares, the global uniform rotation optimal solution is solved. Finally, the iterative strategy is used to eliminate the rough difference and get the exact rotation matrix. The experimental results show that the proposed method is more accurate and more robust than the traditional least square method. The three dimensional reconstruction can get more dense and uniform point cloud, and the reconstruction integrity is better. The accuracy of the rotation average is better than 0.15 degrees, the calculation time is not more than 0.31s, and the reprojection error is less than 1.3 pixels after the beam method adjustment. It basically meets the requirements of fast robust three-dimensional reconstruction.
【作者单位】：信息工程大学;
【基金】：国家自然科学基金资助项目(No.41471387,No.41301526)
【分类号】：O152.5;TP391.41

【相似文献】