CAJ | 학술논문

为从建筑物图像获取三维点云，对运动恢复结构（SFM）和多视角密集匹配（CMVS/PMVS）的三维点云重构进行研究，介绍了从Ladybug3全景相机采集到的图像进行建筑物的三维点云重构过程。首先利用尺度不变特征变换（SIFT）来提取和匹配图像上的特征点并计算多视图之间的几何关系，然后由SFM分析相机运动进而寻找三维点云结构，利用CMVS对图像进行聚簇；最后，采用基于面片模型的PMVS通过匹配、扩展、过滤三个阶段来完成密集匹配同时生成稠密三维点云。实验结果表明，算法能够有效地重构建筑物三维点云，对三维重建有一定的参考价值。
위종건축물도상획취삼유점운，대운동회복결구（SFM）화다시각밀집필배（CMVS/PMVS）적삼유점운중구진행연구，개소료종Ladybug3전경상궤채집도적도상진행건축물적삼유점운중구과정。수선이용척도불변특정변환（SIFT）래제취화필배도상상적특정점병계산다시도지간적궤하관계，연후유SFM분석상궤운동진이심조삼유점운결구，이용CMVS대도상진행취족；최후，채용기우면편모형적PMVS통과필배、확전、과려삼개계단래완성밀집필배동시생성주밀삼유점운。실험결과표명，산법능구유효지중구건축물삼유점운，대삼유중건유일정적삼고개치。
In order to obtain 3D point cloud from building images ,we studied the Structure From Motion (SFM) and Clustering Views for Multi-view Stereo and Patches based Multi-view Stereo (CMVS/PMVS) and introduced the process of 3D point cloud reconstruction of buildings captured from Ladybug3 panoramic camera. Firstly, the Scale-invariant Feature Transform (SIFT) algorithm was used to extract and match the feature points of images , and calculate the geometric relationships between multiple views. Then the camera movement was analyzed by SFM algorithm to acquire 3D point cloud structure and the images were clustered by CMVS algorithm. Finally , the dense matching was accomplished and the dense point cloud was generated by PMVS algorithm which in-volved three steps: matching, expanding and filtering. Experimental results show that the algorithm is able to ef-fectively reconstruct 3D point cloud structure of buildings and highly valuable for 3D reconstruction.