| Evaluation between SVO, ORB SLAM and OKVIS SLAM on Surveying Robot |
| Paper ID : 1311-SMPR |
| Authors: |
|
Ali Arabi *, Mohammad hossein Moazezian, Faeze Najafzadeh, Ali Hosseininaveh, Ali mohammadzadeh Department of Photogrammetry and Remote Sensing Faculty of Geodesy and Geomatics Engineering K. N. Toosi University of Technology |
| Abstract: |
| In recent decade, with grow of use of robots, creating map and positioning in is important for robots performance in unknown indoor environment. The ROS-based robot have different solutions for this problem. This paper present a comparison between three Simultaneous Localization and Mapping (SLAM) method available in Robot Operating System (ROS). For comparison, the SLAM method is tested on surveying robot. Following SLAM techniques is studied: (a) ORB SLAM2, (b) Open Keyframe-based Visual-Inertial SLAM (OKVIS), (c) Semi-direct Visual Odometry (SVO). These techniques needs stereo camera and 2d lidar sensor. Since these method were tested on the surveying robot and kitty dataset, result can be estimated with suitable metric for comparison between mentioned SLAM methods. As the comparison result shows, the OKVIS can create the map and model the trajectory better than ORB SLAM for surveying robot. The accuracy of produced map with OKVIS is 7% better than ORB SLAM and SVO for surveying robot. |
| Keywords: |
| Surveying Robot, ORB SLAM, OKVIS SLAM, SVO, ROS, map |
| Status : Paper Accepted (Oral Presentation) |
