Two-View Structure-from-Motion with Multiple Feature Detector Operators

This paper presents a novel two-view Structure-from-Motion (SfM) algorithm with the application of multiple Feature Detector Operators (FDO). The key this study is implementation FDOs into SfM algorithm. workflow can be divided three general steps: feature detection and matching, pose estimation point cloud (PCL) generation. experimental results, quantitative analyses comparison existing algorithms demonstrate that effectively improve performance Firstly, in Oxford test dataset, RMSE reaches on average 0.11 m (UBC), 0.36 (bikes), 0.52 (trees) 0.37 (Leuven). proves illumination changes, blurring JPEG compression handled satisfactorily. Secondly, EPFL number features lost processes 21% total PCL 27,673 pt, only minimally higher than ORB (20.91%) 10,266 pt. Finally, verification process real-world unmanned aerial vehicle (UAV) shows denser around edges, corners target, speed much faster algorithms. Overall, framework proposed has been proven viable alternative to classical procedure, terms performance, efficiency simplicity.