Laser & Optoelectronics Progress, Volume. 62, Issue 12, 1215005(2025)
Improved Visual SLAM Localization Algorithm for ORB Feature Matching
Figures & Tables(9)
Fig. 2. Feature extraction results. (a) ORB algorithm; (b) MAGSAC++ algorithm; (c) GMS algorithm; (d) light adaptive algorithm; (e) AM-ORB algorithm
Fig. 3. Feature matching effect comparison. (a) Directly matching; (b) ORB algorithm; (c) MAGSAC++ algorithm; (d) GMS algorithm; (e) light adaptive algorithm; (f) AM-ORB algorithm
Table 1. Feature extraction results of different algorithms
View tableTable 1. Feature extraction results of different algorithms
Algorithm Feature extraction time /ms Number of feature points ORB 33.8 450 MAGSAC++ 32.1 452 GMS 30.5 500 Light adaptive 30.9 472 AM-ORB 26.1 500
Table 2. Matching results of different algorithms for the same image
View tableTable 2. Matching results of different algorithms for the same image
Algorithm Matching duration /ms Mismatch removal duration /ms Internal points quantity Matching ratio /% ORB 21.30 7.38 75 37.5 MAGSAC++ 14.72 3.60 117 58.5 GMS 9.35 3.39 135 67.5 Light adaptive 8.29 2.54 138 69.0 AM-ORB 8.11 2.21 145 72.5
Table 3. Comparison results of ATE between our algorithm and ORB-SLAM3
View tableTable 3. Comparison results of ATE between our algorithm and ORB-SLAM3
Dataset RMSE of ATE /m MEAN of ATE /m SSE of ATE /m Std of ATE /m ORB-SLAM3 Ours ORB-SLAM3 Ours ORB-SLAM3 Ours ORB-SLAM3 Ours fr1_xyz 0.009687 0.009396 0.008135 0.007906 0.074128 0.069741 0.005259 0.005078 fr1_desk 0.019704 0.016015 0.015496 0.012844 0.221692 0.146705 0.012170 0.009566 fr2_360_kidnap 0.374350 0.369186 0.177358 0.162416 184.982477 179.9130 0.329670 0.331541 fr2_pioneer_360 0.088865 0.070984 0.078118 0.061499 6.546617 4.177140 0.042362 0.035449 fr3_static 0.008133 0.007549 0.007164 0.006753 0.044719 0.038522 0.003851 0.003373
Table 4. Comparison results of RPE between our algorithm and ORB-SLAM3
View tableTable 4. Comparison results of RPE between our algorithm and ORB-SLAM3
Dateset RMSE of RPE /m MEAN of RPE /m SSE of RPE /m Std of RPE /m ORB-SLAM3 Ours ORB-SLAM3 Ours ORB-SLAM3 Ours ORB-SLAM3 Ours fr1_xyz 0.005840 0.005811 0.004869 0.004823 0.026906 0.026641 0.003225 0.003241 fr1_desk 0.011041 0.009372 0.007880 0.007325 0.069608 0.050157 0.007733 0.005847 fr2_360_kidnap 0.115706 0.117364 0.016027 0.015596 17.658710 18.1684 0.114591 0.116324 fr2_pioneer_360 0.023348 0.020764 0.012678 0.011808 0.451370 0.357004 0.019606 0.017080 fr3_static 0.004980 0.004810 0.004207 0.004134 0.016737 0.015615 0.002664 0.002459
Table 5. Comparison of ATE accuracy of different algorithms
View tableTable 5. Comparison of ATE accuracy of different algorithms
Algorithm RMSE of ATE /m fr1_xyz fr1_desk fr2_desk fr2_pioneer_slam fr2_pioneer_360 fr3_office Ref. ［ 17 ］0.011 0.020 0.009 0.018 Ref. ［ 7 ］0.010 0.019 0.019 0.012 FAST ORB-SLAM［ 18 ］0.010 0.014 0.009 0.011 GCNv2-PROSAC-SLAM［ 19 ］0.019 0.121 0.054 Ours 0.009 0.016 0.005 0.110 0.020 0.011
Tools
Get Citation
Copy Citation Text
Daixian Zhu, Jiaxin Wei, Shulin Liu. Improved Visual SLAM Localization Algorithm for ORB Feature Matching[J]. Laser & Optoelectronics Progress, 2025, 62(12): 1215005
Paper Information
Category: Machine Vision
Received: Oct. 15, 2024
Accepted: Dec. 19, 2024
Published Online: Jun. 9, 2025
The Author Email: Jiaxin Wei (wjx@stu.xust.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20