Laser & Optoelectronics Progress, Volume. 56, Issue 21, 211507(2019)
Optimization of Visual Odometry Algorithm Based on ORB Feature
Figures & Tables(12)
Fig. 8. Triangular networks established by Delauna triangulation. (a) Triangle network of reference frame; (b) triangle network of current frame
Table 1. Root mean square error of camera track
View tableTable 1. Root mean square error of camera track
Image sequence KITTI01 KITTI05 KITTI07 KITTI09 Scene scale /(m×m) 1157×1827 479×426 191×209 465×568 Number of key frames 1101 2761 1101 1591 ORB-SLAM error /m X 8.23 3.36 7.62 Error of proposed algorithm /m 15.66 7.42 3.14 6.58
Table 2. Error table of camera pose estimation
View tableTable 2. Error table of camera pose estimation
Algorithm RMSE /m Mean /m Median /m Min /m Max /m ORB-SLAM 0.087274 0.082128 0.075679 0.021258 0.164835 Proposed algorithm 0.060309 0.062565 0.067717 0.024182 0.163003
Table 3. Camera motion errors in different scenes
View tableTable 3. Camera motion errors in different scenes
Algorithm station Average error of improved ORB Average error of standard ORB Reduction ratio of average error Average distance error /m Average angle error /(°) Average distance error /m Average angle error /(°) Reduction ratio of distance error /m Reduction ratio of angle error /(°) Scene 1 0.08 2.1 0.71 11.8 88.73 82.20 Scene 2 0.13 4.2 0.95 18.3 86.32 77.05 Scene 3 0.52 9.3 1.66 20.6 68.67 54.84
Tools
Get Citation
Copy Citation Text
Fuchun Lin, Yuhong Liu, Jinfan Zhou, Zhinan Ma, Qianqian He, Manman Wang, Rongfen Zhang. Optimization of Visual Odometry Algorithm Based on ORB Feature[J]. Laser & Optoelectronics Progress, 2019, 56(21): 211507
Paper Information
Category: Machine Vision
Received: Apr. 3, 2019
Accepted: May. 6, 2019
Published Online: Nov. 2, 2019
The Author Email: Rongfen Zhang (rfzhang@gzu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20