Evaluation and Software Implementations in Out-/Indoor Navigation based on Visual Odometry for Mono Cameras

Status: Abgeschlossen, April 2016

Bearbeitung: M.Sc. Fickrie Muhammad

The navigation solution for indoor environments especially for GPS-denied area is evaluated in this thesis, based on a monocular web camera without artificial markers or IMU (inertial measurement unit) sensors. The evaluation mainly focus on the feature tracking robustness of  PTAM  and  SVO  algorithms.

The evaluation approach considers the drift of SVO and PTAM against  ground  truth,  which  the  least  squares  estimation is used to estimate scale, rotation and translation between SVO, PTAM and ground truth frames. The projection error of camera is calculated with camera calibration software from ROS and PTAM package.

Position drift  (a) in x-axis, (b) in y-axis, (c) in z-axis of PTAM and SVO comparing with Ar tags

Video link on youtube

HSKA-Link der Zusammenfassung