基于MPB的三维重建方法研究及实现

摘要

This paper investigated the 3D reconstruction of the UAV remote sensing images,and prpopsed a novel reconstruc-tion framework(MPB)which explicitly decomposed the reconstruction task into image mosaic,camera pose estimation and bundle adjustment.Image mosaic means that all remote sensing images would be registrated into a plane to establish the plane model of terrain,then it estimated the camera pose from the plane model of terrain,bundle adjustment the camera poses and the terrain coordinates to complete the terrain 3D sparse reconstruction.This paper got the dense points based on the PMVS algo-rithm,and did a surface reconstruction and texture mapping to complete the 3D reconstruction.Based on the above ideas,it de-velopped a UAV remote sensing image 3D reconstruction system,and conducted extensive testing.The results show that the sys-tem can improve the reconstruction efficiency greatly,providing a fast and efficient approach for 3D reconstruction of remote sensing data.%针对无人机遥感影像三维重建需求，提出了基于MPB（mosaic-pose-bundle）的三维重建方法，将三维重建过程分为影像拼接、姿态估计、捆集调整三个步骤。影像拼接是将所有遥感影像配准到一个基准平面，建立地形的平面模型；姿态估计则根据平面模型解算出相机的外参数，完成相机的概率定向；捆集调整则是对定向参数和平面坐标进行非线性优化，完成地形的稀疏三维重建，并在此基础上重建场景的PMVS点云，进行表面重建及纹理映射，完成地形的实景真三维重建。依据上述思路，开发了一套基于无人机遥感影像的三维重建系统，通过对不同地形的遥感影像进行测试，验证了方法的正确性及稳定性，相对于传统的三维重建系统更加高效，为遥感影像的三维重建提供一种高效、稳健的新方法。