Laser & Optoelectronics Progress, Volume. 62, Issue 10, 1015013(2025)
Binocular-Vision SLAM Based on Improved ELSED Line-Detection Algorithm
Figures & Tables(13)
Fig. 3. Comparison experiment of line feature detection carried out on the EuRoC dataset. (a) Original ELSED detection line feature effect; (b) improved ELSED detection line feature effect
Fig. 4. Comparison experiment of line feature matching optimization carried out on the EuRoC dataset. (a) Unoptimized matching effect; (b) optimized matching effect
Fig. 5. Comparison of trajectory evaluation results for different algorithms. (a) ORB-SLAM3; (b) ORB-LINE-SLAM; (c) ORB&ELSED-SLAM(O); (d) ORB&ELSED-SLAM(M)
Fig. 6. Comparison between trajectories of 01, 02, 08, and 09 sequences predicted by ORB&ELSED(M) and real trajectories. (a) 01 sequence trajectory; (b) 02 sequence trajectory; (c) 08 sequence trajectory; (d) 09 sequence trajectory
Fig. 7. Comparison of predicted trajectories and true trajectories of four algorithms on fantasy-csc1 sequence. (a) ORB-SLAM3; (b) ORB-LINE-SLAM; (c) ORB&ELSED-SLAM(O); (d) ORB&ELSED-SLAM(M)
Table 1. Performance comparison of line feature extraction algorithms
View tableTable 1. Performance comparison of line feature extraction algorithms
Sequence ELSED（M） ELSED（O） LSD Time /ms Line number Length /pixel Time /ms Line number Length /pixel Time /ms Line number Length /pixel MH01 32.4 90 87.44 36.0 179 68.77 141.7 174 53.01 MH02 31.2 87 89.90 35.3 168 70.62 127.0 168 53.69 MH03 33.1 86 90.76 37.5 166 70.73 132.3 165 54.93 MH04 23.8 89 97.61 27.3 157 76.57 106.3 163 57.95 MH05 24.1 87 96.81 28.1 156 75.51 107.2 160 57.56
Table 2. Performance comparison of line feature matching algorithms
View tableTable 2. Performance comparison of line feature matching algorithms
Sequence LBD Optimized LBD Number of matches Success rate /% Number of matches Success rate /% MH01 74 80.92 67 88.62 MH02 76 81.44 66 87.31 MH03 71 83.61 67 89.62 MH04 75 83.21 69 87.23 MH05 71 82.53 65 90.51
Table 3. Performance comparison of different algorithms on the EuRoC dataset
View tableTable 3. Performance comparison of different algorithms on the EuRoC dataset
Sequence ORB-SLAM2 ORB-SLAM3 PL-SLAM ORB_LINE_SLAM RTE ATE RTE ATE RTE ATE RTE ATE MH01 2.2345 0.0363 2.6345 0.0349 2.5677 0.0416 2.2378 0.0381 MH02 2.2269 0.0373 2.3269 0.0418 2.3422 0.0522 — — MH03 5.1396 0.0379 5.2396 0.0471 5.4144 0.0399 5.1404 0.0401 MH04 5.9624 0.1002 6.0624 0.0654 6.0132 0.0641 5.9704 0.0449 MH05 5.4049 0.0506 5.4049 0.0539 5.6039 0.0697 5.4032 0.0452 V101 2.3503 0.0868 2.4503 0.0882 2.3143 0.0923 — — V102 3.4181 0.0616 3.4181 0.0702 3.2145 0.0659 — — V103 2.2320 0.0820 2.3320 0.1418 2.4132 0.0889 — — V201 1.9059 0.0668 1.8666 0.0777 2.0671 0.0709 1.8699 0.0619 V202 3.0680 0.0628 3.1680 0.0938 3.2103 0.0565 3.0799 0.0549 V203 3.2517 0.2393 3.2517 0.7727 3.4351 0.3261 — — Sequence Ref. ［ 11 ］ algorithmORB-LINE-SLAM ORB&ELSED（O） ORB&ELSED（M） RTE ATE RTE ATE RTE ATE RTE ATE MH01 1.9842 0.0361 2.2371 0.0372 2.1344 0.0369 2.0671 0.0366 MH02 2.2277 0.0398 2.0239 0.0378 2.2311 0.0368 2.0133 0.0359 MH03 5.1233 0.0411 5.2338 0.0381 5.1133 0.0393 5.0971 0.0368 MH04 5.9231 0.0433 5.9744 0.1038 5.9711 0.0995 5.8966 0.0607 MH05 5.1022 0.0463 5.3979 0.0485 5.3413 0.0481 5.0631 0.0443 V101 2.2157 0.0855 2.3521 0.0868 2.2566 0.0867 2.4134 0.0849 V102 3.8971 0.0622 3.4114 0.0615 3.3566 0.0601 3.4566 0.0599 V103 2.0671 0.0765 2.2779 0.0877 2.3712 0.0784 2.1433 0.0882 V201 1.9834 0.0655 1.8715 0.0627 1.9633 0.0662 1.7652 0.0581 V202 3.1544 0.0574 3.0837 0.0599 3.1752 0.0596 3.0211 0.0543 V203 3.3471 0.2234 3.7847 0.2641 3.3274 0.2937 3.1752 0.1909
Table 4. Performance comparison of different algorithms on KITTI dataset
View tableTable 4. Performance comparison of different algorithms on KITTI dataset
Sequence ORB-LINE-SLAM ORB&ELSED（O） ORB&ELSED（M） RTE ATE RTE ATE RTE ATE 00 0.0282 1.3074 0.0271 1.3227 0.0233 1.2706 01 0.0578 10.0760 0.0514 10.4470 0.0465 9.3312 02 0.0292 6.3464 0.0275 6.7165 0.0214 5.6930 03 0.0171 0.7817 0.0173 0.7273 0.0186 0.7845 04 0.0203 0.9920 0.0194 0.9896 0.0185 0.8000 05 0.0172 0.7548 0.0169 0.7709 0.0154 0.7215 06 0.0185 0.5899 0.0174 0.5954 0.0165 0.5057 07 0.0168 0.4943 0.0163 0.4977 0.0168 0.4652 08 0.0385 3.4890 0.0364 3.4017 0.0354 3.2581 09 0.0210 1.6675 0.0207 1.6847 0.0201 1.6114 10 0.0199 1.0869 0.0199 1.1277 0.0199 1.0264
Table 5. Performance comparison of different algorithms on UMA dataset
View tableTable 5. Performance comparison of different algorithms on UMA dataset
Sequence ORB-SLAM3 Ref. ［ 11 ］ algorithmORB-LINE-SLAM ORB&ELSED（O） ORB&ELSED（M） RTE ATE RTE ATE RTE ATE RTE ATE RTE ATE Corridor-eng — — 0.1537 0.1521 0.1603 0.1623 0.1757 0.1673 0.1444 0.1481 Hall1-rev-eng 0.2791 0.2888 0.2133 0.1033 0.2183 0.0617 0.2486 0.0621 0.2087 0.0604 Class-csc1 — — 0.2713 0.0671 0.2887 0.0517 0.2933 0.0622 0.2896 0.0764 Class-csc2 — — 0.3457 0.0492 0.3384 0.0518 0.3776 0.0745 0.3158 0.0509 Conference-csc1 0.0810 0.2152 0.1035 0.0243 0.1251 0.0279 0.1229 0.0217 0.1220 0.0208 Fantasy-csc1 0.2371 0.0909 0.2089 0.0513 0.2268 0.0598 0.2318 0.0677 0.2045 0.0367
Table 6. Running time of different algorithms
View tableTable 6. Running time of different algorithms
Sequence Ref. ［ 11 ］ algorithmORB-LINE-SLAM ORB&ELSED（O） ORB&ELSED（M） Corridor-eng 191 269 209 185 Hall1-rev-eng 206 271 216 200 Class-csc1 201 304 207 206 Class-csc2 197 294 221 204 Conference-csc1 233 290 246 223 Fantasy-csc1 203 211 202 196
Tools
Get Citation
Copy Citation Text
Xu Zhu, Xiping Xu, Ning Zhang, Zhi Meng, Hongwei Tan, Luqing Zhang, Yiming Hu, Jiaxu Zhang. Binocular-Vision SLAM Based on Improved ELSED Line-Detection Algorithm[J]. Laser & Optoelectronics Progress, 2025, 62(10): 1015013
Paper Information
Category: Machine Vision
Received: Aug. 13, 2024
Accepted: Dec. 2, 2024
Published Online: Apr. 28, 2025
The Author Email: Xiping Xu (xxp@cust.edu.cn)
CSTR:32186.14.LOP241835
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20