Visual SLAM optimization algorithm based on dynamic object detection

序列	ORB-SLAM2					RDFP-SLAM					提升
序列	Max/m	Mean/m	Median/m	Min/m	RMES/m	Max/m	Mean/m	Median/m	Min/m	RMES/m	RMSE /%
s_xyz	0.0457	0.0100	0.0081	0.0020	0.0126	0.0513	0.0117	0.0105	0.0026	0.0134	−6.34
s_rpy	0.0797	0.0223	0.0205	0.0050	0.0248	0.0723	0.0200	0.0179	0.0046	0.0246	0.81
s_static	0.0166	0.0102	0.0107	0.0040	0.0110	0.0067	0.0040	0.0042	0.0010	0.0043	60.91
w_xyz	0.5732	0.2421	0.2262	0.0478	0.2700	0.0502	0.0119	0.0104	0.0030	0.0138	94.89
w_static	0.0666	0.0132	0.0098	0.0010	0.0174	0.0233	0.0071	0.0065	0.0012	0.0060	54.54
w_halfsphere	0.5901	0.3552	0.4036	0.1003	0.3724	0.0917	0.0203	0.0159	0.0027	0.0258	93.07
01	11.1821	5.3211	5.7516	0.6635	5.7513	9.9161	4.6459	4.8107	0.5486	5.1014	11.30
04	0.3679	0.1630	0.1646	0.0317	0.1809	0.3258	0.1475	0.1467	0.0227	0.1677	7.30
05	1.2044	0.5371	0.5364	0.0667	0.5820	1.0281	0.4341	0.4067	0.0258	0.4638	20.31
08	23.1262	4.9016	4.4253	0.3446	6.0353	20.4319	4.8773	4.0426	0.1695	5.6029	7.16
09	17.5842	7.2730	5.4687	2.3194	8.6998	16.0978	5.9493	5.1122	2.0694	7.9495	8.62

Table 2. Comparison of relative trajectory errors between ORB-SLAM2 algorithm and proposed RDFP-SLAM algorithm

View table

View in Article

Table 2. Comparison of relative trajectory errors between ORB-SLAM2 algorithm and proposed RDFP-SLAM algorithm

序列	ORB-SLAM2					RDFP-SLAM					提升
序列	Max/m	Mean/m	Median/m	Min/m	RMES/m	Max/m	Mean/m	Median/m	Min/m	RMES/m	RMSE /%
s_xyz	0.0521	0.0146	0.0112	0.0016	0.0184	0.0675	0.0152	0.0122	0.0033	0.0186	−1.08
s_rpy	0.0738	0.0153	0.0132	0.0021	0.0194	0.1017	0.0179	0.0114	0.0010	0.0209	−7.73
s_static	0.0196	0.0130	0.0144	0.0043	0.0139	0.0099	0.0064	0.0062	0.0017	0.0069	50.36
w_xyz	0.1975	0.0154	0.0119	0.0005	0.0226	0.0470	0.0140	0.0117	0.0011	0.0171	24.34
w_static	0.0143	0.0017	0.0011	0.0003	0.0027	0.0133	0.0011	0.0005	0.0010	0.0023	14.81
w_halfsphere	0.4718	0.0236	0.0160	0.0024	0.0475	0.1050	0.0186	0.0137	0.0017	0.0248	47.79
01	5.3284	2.9760	3.1079	1.6564	3.1642	5.2179	2.9157	2.9760	1.6418	3.0523	3.54
04	1.3549	1.2802	1.2802	1.2056	1.2826	1.2181	1.1540	1.1515	1.0098	1.1605	9.52
05	2.0658	0.8591	0.9124	0.0489	0.9600	2.0122	0.7741	0.8106	0.0415	0.8521	11.24
08	17.4345	2.4349	2.0597	0.6849	3.5244	17.3589	2.2462	1.8145	0.4548	3.3170	5.88
09	4.5879	2.2908	1.8768	0.8455	2.5270	4.1145	2.0317	1.6074	0.6765	2.2901	9.37

Table 3. Comparison of absolute trajectory errors of different algorithms in TUM dataset m

View table

View in Article

Table 3. Comparison of absolute trajectory errors of different algorithms in TUM dataset m

序列	ORB-SLAM2	DS-SLAM	DynaSLAM	Semantic-SLAM	文献[15]	RDFP-SLAM
序列	RMSE	RMSE	RMSE	RMSE	RMSE	RMSE
s_xyz	0.0126	0.0108	0.0140	0.0143	0.0104	0.0124
s_static	0.0110	0.0065	0.0067	0.0081	0.0058	0.0043
w_xyz	0.2700	0.0247	0.0150	0.0225	0.0150	0.0138
w_rpy	0.5333	0.4442	0.0400	0.3641	---	0.0436
w_static	0.0174	0.0081	0.0090	0.0096	0.0066	0.0060
w_halfsphere	0.3724	0.0303	0.0270	0.0384	0.0283	0.0258

Table 4. Comparison of relative trajectory errors of different algorithms in KITTI dataset %
View table
View in Article
Table 4. Comparison of relative trajectory errors of different algorithms in KITTI dataset %
序列 ORB-SLAM2 DynaSLAM 文献[14] RDFP-SLAM
RMSE RMSE RMSE RMSE
01 3.1079 3.2721 3.0470 2.9760
04 1.2802 1.1845 1.1904 1.1515
05 0.9124 0.8311 0.8925 0.8106
08 2.0597 1.9674 1.9463 1.8145
09 1.8768 1.6943 1.7121 1.6074

Table 5. Comparison of absolute trajectory errors of different algorithms in real environments m
View table
View in Article
Table 5. Comparison of absolute trajectory errors of different algorithms in real environments m
序列 ORB-SLAM2 DS-SLAM DynaSLAM 文献[14] RDFP-SLAM
RMSE RMSE RMSE RMSE RMSE
车间环境1（RGB-D） 0.4253 0.0402 0.0289 --- 0.0225
车间环境1（Stereo） 0.4012 --- 0.026 7 0.0274 0.0168

Table 6. Comparison of average tracking time of different algorithms in real environments s

View table

View in Article

Table 6. Comparison of average tracking time of different algorithms in real environments s

序列	ORB-SLAM2	DS-SLAM	DynaSLAM	文献[14]	RDFP-SLAM
序列	RMSE	RMSE	RMSE	RMSE	RMSE
车间环境1（RGB-D）	0.0220	0.1786	0.3465	---	0.1126
车间环境2（RGB-D）	0.0217	0.1732	0.2843	---	0.1087
车间环境1（Stereo）	0.0534	---	0.354 9	0.1192	0.1267
车间环境2（Stereo）	0.0601	---	0.365 6	0.1301	0.1346

Tools

Get Citation

Copy Citation Text

Xubiyue SHANG, Junwei TIAN, Xingang WANG, Yuhan BU, Wenbo ZHANG. Visual SLAM optimization algorithm based on dynamic object detection[J]. Journal of Applied Optics, 2025, 46(1): 102

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category:

Received: Oct. 26, 2023

Accepted: --

Published Online: Apr. 1, 2025

The Author Email: Junwei TIAN (田军委)

DOI:10.5768/JAO202546.0102002

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Comparison of absolute trajectory errors between ORB-SLAM2 algorithm and proposed RDFP-SLAM algorithm

Table 1. Comparison of absolute trajectory errors between ORB-SLAM2 algorithm and proposed RDFP-SLAM algorithm

Table 2. Comparison of relative trajectory errors between ORB-SLAM2 algorithm and proposed RDFP-SLAM algorithm

Table 2. Comparison of relative trajectory errors between ORB-SLAM2 algorithm and proposed RDFP-SLAM algorithm

Table 3. Comparison of absolute trajectory errors of different algorithms in TUM dataset m

Table 3. Comparison of absolute trajectory errors of different algorithms in TUM dataset m

Table 4. Comparison of relative trajectory errors of different algorithms in KITTI dataset %

Table 4. Comparison of relative trajectory errors of different algorithms in KITTI dataset %

Table 5. Comparison of absolute trajectory errors of different algorithms in real environments m

Table 5. Comparison of absolute trajectory errors of different algorithms in real environments m

Table 6. Comparison of average tracking time of different algorithms in real environments s

Table 6. Comparison of average tracking time of different algorithms in real environments s