| Title |
Visual Inertial Navigation System Optimization using The Depth of Features with Stereo |
| Authors |
박준수(Jun-su Park) ; 박순용(Soon-Yong Park) |
| DOI |
https://doi.org/10.5573/ieie.2021.58.7.77 |
| Keywords |
SLAM; Camera; IMU; VINS; Stereo; Fusion |
| Abstract |
Visual-Inertial SLAM combines low cost camera and inertial measurement sensors and can achievers high performance tracking using the advantages of image features and fast motion estimation of inertial sensor. Recently, there are increasing applications of VI-SLAM in many areas such as robotics, augmented reality, autonomous vehicle. In the VINS-mono technique which employing a single camera, the feature tracking in consecutive image frames affects the camera motion estimation because the depth of image features should be solved in the state vector optimization. If image features fail to track, it is subject to error in the state optimization of VINS-mono which uses a single camera. This paper proposes a VINS-stereo technique in order to estimate accurate camera motion even in fast camera motion. The depth of image features is determined by stereo matching. We add a cost function of feature depth to the state vector optimization step and accurately estimate the camera motion. |