利用无人机影像和广义迭代重加权最小二乘法的三维重建

摘要

In order to improve the computing efficiency of 3D reconstruction of UAV (unmanned aerial vehicle) images,a global reconstruction method for UAV images applying GIRLS (generalized iteratively reweighted least squares) was proposed.First of all,the basic idea and concrete steps of GIRLS were introduced.Then after converting the problem of solving global rotation matrixes to solving rotation vectors in Lie algebra,the optimal solution of global rotation was obtained using an algorithm combining norm optimization and IRLS (iteratively reweighted least squares).And afterwards,according to the epipolar geometry constraint,the relative translation direction vectors among cameras were calculated in the global coordinate system,and the optimal solution of global position was acquired by means of quadratic programming within the framework of GIRLS.Finally,the position and orientation parameters were refined with the reconstructed 3D points by bundle adjustment.Experimental results indicated that the proposed method could recover three-dimension scene better while improving the reconstruction efficiency.%为了提高基于无人机影像重建三维场景的效率,提出了应用广义迭代重加权最小二乘法的全局式三维重建方法.首先,介绍了广义迭代重加权最小二乘法的基本思想和具体步骤;接着,将全局旋转矩阵的求解转化为李代数中旋转向量的求解问题,利用结合范数优化和迭代重加权最小二乘的方法求出全局旋转最优解;然后,根据极线几何约束条件,解算相机在全局坐标系下的相对平移方向向量,在广义迭代重加权最小二乘架构下,利用二次规划法求出相机全局位置最优解;最后,对影像位姿参数和重构三维点进行光束法平差优化.实验结果表明该方法在提高重建效率的同时,能够更真实地恢复场景的几何形态.