基于单目增量式稀疏SFM研究与实现  

Research and Implementation of Monocular Sparse Reconstruction Based on Incremental SFM

作　　者：安帅 杨杰[1] 乔羽 李鹏[1] AN Shuai;YANG Jie;Qiao Yu;Li Peng(School of Electrical Engineering,Qingdao University,Qingdao 266071,China)

机构地区：[1]青岛大学机电工程学院,青岛266071

出　　处：《青岛大学学报（自然科学版）》2018年第2期102-107,共6页Journal of Qingdao University(Natural Science Edition)

摘　　要：为研究视觉SLAM(Simutaneous Localization and Mapping)以及视觉SFM(Structure From Motion)领域中机器人能低成本重建出可进行语义识别三维场景的问题,借鉴Colmap算法,对重建过程中的部分步骤加以改进,对初始化图像的选择进行了精选,在重建出有误差的三维点云中进行多次基于光束平差法的优化,对无人机拍摄的图像进行重建分析,结果表明,图像在重建的完整度、鲁棒性、精确度以及效率等方面,均取得了较好的效果。In the field of visual SLAM and visual SFM,in order to study the problem of reconstructing the3 Dscene that can be semantic recognized by the robot at a low cost,some of the procedure of the reconstruction has been improved based on the Colmap algorithm.Photos in the initialization process have been carefully selected and Bundle Adjustment after the 3 Dpoint cloud has been computed several times.Reconstruction result has been analyzed by using photos pictured by a UAV,and it has been found that the completeness,robustness,accuracy and the efficiency of the reconstruction has received a good result.

关 键 词：稀疏重建 增量SFM 运动恢复结构 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]