Visual SLAM Algorithm Combining Image Brightness Enhancement Module and IMU Information

Fig. 4. Comparison of results of feature points extracted by two algorithms. (a) Extraction result of ORB-SLAM2 algorithm; (b) extraction result of AGTFIS-SLAM algorithm

Download full size

Fig. 5. Schematic diagram of LK optical flow method

Download full size

Fig. 6. Flowchart of AGTFIS-SLAM algorithm

Download full size

Fig. 7. Comparison of maps and trajectories of two algorithms. (a) ORB-SLAM2 algorithm map; (b) AGTFIS-SLAM algorithm map; (c) comparison between estimated trajectory of ORB-SLAM2 algorithm and real trajectory; (d) comparison between estimated trajectory of AGTFIS-SLAM algorithm and real trajectory; (e) comparison of precision and recall curves of two algorithms

Download full size

Fig. 8. Image sequence with dimly processing

Download full size

Fig. 9. Comparison of absolute trajectory errors of three algorithms

Download full size

Fig. 10. Comparison of relative pose errors of three algorithms

Download full size

Fig. 11. Comparison of average tracking time per frame between two algorithms

Download full size

Fig. 12. Bingda ROS robot

Download full size

Fig. 13. Laboratory environment. (a) Before treatment; (b) after treatment

Download full size

Fig. 14. Verification results of AGTFIS-SLAM algorithm. (a) Feature point extraction result; (b) key frame generation result

Download full size

Table 1. Maximum grayscale values enhanced of algorithm by influence of $u$ value when $γ$ =1
View table
Table 1. Maximum grayscale values enhanced of algorithm by influence of $u$ value when $γ$ =1
$u$ Maximum grayscale value
0.1 36
0.2 51
0.3 77
0.4 102
0.5 128
0.6 153
0.7 179
0.8 204
0.9 230

Table 2. Tracking accuracies of AGTFIS-SLAM algorithm with different $u$ values
View table
Table 2. Tracking accuracies of AGTFIS-SLAM algorithm with different $u$ values
$u$ ATT /ms ATE /m RPE /m
$u$ =0.3 37 0.0541 0.6112
$u$ =0.4 40 0.0492 0.6351
$u$ =0.5 36 0.0506 0.6199
$u$ =0.6 42 0.0513 0.6149
$u$ =0.7 44 0.0523 0.6197

Table 3. Absolute trajectory errors

View table

Table 3. Absolute trajectory errors

Data sequence	Only image enhancement	Only IMU	ORB-SLAM2	AGTFIS- SLAM
MH_04_difficult	0.0566	0.0574	0.0580	0.0560
MH_05_difficult	0.0532	0.0828	0.1265	0.0506
V2_03_difficult	0.1658	0.2257	0.3102	0.1535
Average	0.0919	0.1220	0.1649	0.0867

Table 4. Absolute trajectory errors

View table

Table 4. Absolute trajectory errors

Data sequence	ORB- SLAM2	ORB- SLAM3	AGTFIS- SLAM
fr1.xyz	0.0550	0.0242	0.0120
fr1.desk	—	0.0303	0.0144
fr1.room	—	—	0.0986
MH_04_difficult	0.0580	0.0508	0.0560
MH_05_difficult	0.1265	0.0536	0.0506
V1_01_easy	0.0953	0.0957	0.0950
V1_02_medium	0.0640	0.0630	0.0649
V1_03_difficult	0.0892	0.0664	0.0639
V2_01_easy	0.0591	0.0732	0.0575
V2_02_medium	0.0567	0.0561	0.0546
V2_03_difficult	0.3102	0.1674	0.1535
Average	0.1015	0.0681	0.0655

Table 5. Relative pose errors

View table

Table 5. Relative pose errors

Data sequence	ORB- SLAM2	ORB- SLAM3	AGTFIS- SLAM
fr1.xyz	0.0379	0.0189	0.0140
fr1.desk	—	0.0207	0.0160
fr1.room	—	—	0.0597
MH_04_difficult	0.6429	0.5389	0.6076
MH_05_difficult	0.6203	0.4471	0.4271
V1_01_easy	0.5885	0.4883	0.5160
V1_02_medium	0.5801	0.5078	0.5658
V1_03_difficult	0.5355	0.4153	0.4855
V2_01_easy	0.2345	0.1820	0.2032
V2_02_medium	0.5334	0.4764	0.5242
V2_03_difficult	0.4720	0.4238	0.4743
Average	0.4717	0.3519	0.3539

Table 6. Average tracking time per frame

View table

Table 6. Average tracking time per frame

Data sequence	ORB-SLAM2	AGTFIS-SLAM
fr1.xyz	33	21
fr1.desk	—	23
fr1.room	—	29
MH_04_difficult	37	25
MH_05_difficult	41	36
V1_01_easy	34	31
V1_02_medium	34	27
V1_03_difficult	30	27
V2_01_easy	42	32
V2_02_medium	47	33
V2_03_difficult	44	31
Average	38	29

Tools

Get Citation

Copy Citation Text

Bo Wang, Danfeng Shen, Pengfei Bai. Visual SLAM Algorithm Combining Image Brightness Enhancement Module and IMU Information[J]. Laser & Optoelectronics Progress, 2024, 61(24): 2428013

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites