Laser & Optoelectronics Progress, Volume. 60, Issue 4, 0410014(2023)
Laser Simultaneous Localization and Mapping Algorithm Based on Adaptive Features and Closed-Loop Optimization
Figures & Tables(17)
Fig. 3. Comparison of feature extraction at different distances. (a) (b) (c) Original feature extraction effects at distances of 5, 15, 30 m respectively; (d) (e) (f) effects of proposed feature extraction at distances of 5, 15, 30 m respectively
Fig. 4. Effect maps of filtering feature points of ground point cloud. (a) Before filtering out ground feature points;(b) after filtering out ground feature points
Fig. 10. Global point cloud map comparison on KITTI 27. (a) Proposed algorithm; (b) LIO-SAM; (c) LEGO-LOAM; (d) A-LOAM
Fig. 11. Absolute/relative pose error global distribution. (a) Proposed algorithm; (b) LIO-SAM; (c) LEGO-LOAM; (d) A-LOAM
Fig. 13. Comparison of mapping of each algorithm on campus dataset. (a) Proposed algorithm; (b) LIO-SAM; (c) LEGO-LOAM; (d) A-LOAM
Table 1. 6-DOF pose deviation of odometry
View tableTable 1. 6-DOF pose deviation of odometry
Dataset Algorithm X /m Y /m Z /m Roll /（°） Pitch /（°） Yaw /（°） Park Proposed algorithm 1.81 0.23 0.92 1.65 0.91 0.32 LIO-SAM 1.12 0.31 1.86 1.77 0.86 0.28 LEGO-LOAM 2.35 0.48 1.64 1.35 0.93 0.47 A-LOAM 2.79 0.35 2.97 2.89 2.36 0.89 KITTI 27 Proposed algorithm 0.38 0.57 0.98 0.12 0.38 0.23 LIO-SAM 0.26 0.53 1.34 0.25 0.31 0.27 LEGO-LOAM 0.49 0.78 2.06 0.86 0.68 0.38 A-LOAM 0.62 0.77 2.68 0.97 0.54 0.52
Table 2. Absolute pose error and relative pose error
View tableTable 2. Absolute pose error and relative pose error
Dataset Algorithm Absolute pose error Relative pose error Max Mean Min Max Mean Park LIO-SAM 0.76 0.33 0.04 0.38 0.18 A-LOAM 3.62 1.33 0.10 0.47 0.25 LEGO-LOAM 1.69 0.67 0.05 0.42 0.23 Proposed algorithm 0.63 0.25 0.01 0.33 0.12 Proposed algorithm（without Better Feature） 0.92 0.39 0.06 0.39 0.21 Proposed algorithm（without SC-ICP） 3.35 1.08 0.08 0.56 0.27 KITTI 27 LIO-SAM 1.54 0.52 0.15 1.82 0.63 A-LOAM 4.00 1.38 0.10 4.59 1.86 LEGO-LOAM 3.98 1.16 0.01 3.80 1.59 Proposed algorithm 1.66 0.55 0.03 1.68 0.52 Proposed algorithm（without Better Feature） 1.94 0.75 0.11 2.07 0.84 Proposed algorithm（without SC-ICP） 3.54 1.02 0.16 4.21 1.77
Table 3. Comparison of the processing speed of each algorithm
View tableTable 3. Comparison of the processing speed of each algorithm
Algorithm Feature Extraction Scan matching Loop closure Odometry Mapping Proposed algorithm Planar：807 Corner：321 1.94 35.38 8.22 11.53 91.85 Proposed algorithm（without Better Feature） Planar：1547 Corner：535 8.36 49.34 10.35 21.4 114.38 Proposed algorithm（without SC-ICP） Planar：884 Corner：381 2.34 40.75 15.36 15.73 109.24 LIO-SAM Planar：1479 Corner：411 4.21 63.26 12.19 18.77 123.50 LEGO-LOAM Planar：573 Corner：186 1.37 38.47 11.36 16.82 86.57 A-LOAM Planar：1135 Corner：386 35.54 98.64 41.63 215.80
Table 4. Odometry error and pressure test
View tableTable 4. Odometry error and pressure test
Algorithm Odometry /m Odometry error /m Max rate Proposed algorithm 892.8 0.7 15 LIO-SAM 894.6 1.1 15 LEGO-LOAM 891.4 2.1 5 A-LOAM 896.4 2.9 3
Tools
Get Citation
Copy Citation Text
Hejun Wei, Enyong Xu, Bing Han, Yanmei Meng, Jin Wei, Zhengqiang Li. Laser Simultaneous Localization and Mapping Algorithm Based on Adaptive Features and Closed-Loop Optimization[J]. Laser & Optoelectronics Progress, 2023, 60(4): 0410014
Paper Information
Category: Image Processing
Received: Nov. 26, 2021
Accepted: Jan. 5, 2022
Published Online: Feb. 13, 2023
The Author Email: Yanmei Meng (gxu_mengyun@163.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20