HRegNet-LO: LiDAR Odometry Measurement Based on End-to-End Deep Neural Network

Yongjian Fu; Zongchun Li; Hua He; Li Wang; Cong Li

doi:10.3788/AOS230548

Acta Optica Sinica, Volume. 43, Issue 24, 2428003(2023)

HRegNet-LO: LiDAR Odometry Measurement Based on End-to-End Deep Neural Network

Yongjian Fu, Zongchun Li^*, Hua He, Li Wang, and Cong Li

Institute of Geospatial Information, PLA Strategic Support Force Information Engineering University, Zhengzhou 450001, Henan , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(11)

Fig. 1. Flow chart of HRegNet-LO algorithm

Download full size

Fig. 2. Feature points extraction result of a point cloud

Download full size

Fig. 3. Flow chart of HRegNet registration network^[19]

Download full size

Fig. 4. Illustration of feature points matching. (a) Point set ${}_{T_{1}^{n}} F_{n}$ ; (b) point set $F_{m a p}$

Download full size

Fig. 5. Visualization of results of proposed algorithm on testing datasets along each coordinate axis. (a) Seq 08; (b) Seq 09; (c) Seq 10

Download full size

Fig. 6. Aerial view of results of proposed algorithm on testing datasets. (a) Seq 08; (b) Seq 09; (c) Seq 10

Download full size

Fig. 7. Fixed distance relative errors of three algorithms on testing datasets. (a) Seq 08; (b) Seq 09; (c) Seq 10

Download full size

Fig. 8. Fixed distance relative errors of algorithms in ablation experiments. (a) Seq 08; (b) Seq 09; (c) Seq 10

Download full size

Table 1. Details of experimental datasets

View table

Table 1. Details of experimental datasets

Seq	Frame	Length /m	Max speed /（km·h^-1）	Seq	Frame	Length /m	Max speed /（km·h^-1）
00	4541	3724	46	06	1101	1232	51
01	1101	2453	96	07	1101	694	39
02	4661	5067	49	08	4071	3222	43
03	801	560	31	09	1591	1705	52
04	271	393	56	10	1201	919	51
05	2761	2205	40

Table 2. Accuracy and time consuming of three algorithms on testing datasets

View table

Table 2. Accuracy and time consuming of three algorithms on testing datasets

Algorithm	Accuracy and time consuming	Seq 08	Seq 09	Seq 10	RMSE
LOAM	RRE /［（°）·m^-1］	0.0041	0.0031	0.0032	0.0035
	RTE /%	1.0569	0.7326	1.0706	0.9661
	Time consuming /ms	87	83	85	—
F-LOAM	RRE /［（°）·m^-1］	0.0048	0.0051	0.0060	0.0053
	RTE /%	1.3532	1.2520	1.5218	1.3802
	Time consuming /ms	76	74	75	—
HRegNet-LO	RRE /［（°）·m^-1］	0.0033	0.0031	0.0039	0.0035
	RTE /%	1.0472	0.7190	1.0480	0.9508
	Time consuming /ms	95	97	96	—

Table 3. Results of ablation experiments

View table

Table 3. Results of ablation experiments

Algorithm	Accuracy	Seq 08	Seq 09	Seq 10	RMSE
ICP-odom	RRE /［（°）.m^-1］	0.0621	0.1177	0.0586	0.0840
ICP-odom	RTE /%	14.5136	31.0251	16.7450	22.0121
HRegNet-odom	RRE /［（°）.m^-1］	0.0402	0.0161	0.0212	0.0278
HRegNet-odom	RTE /%	9.8344	3.7263	4.5524	6.6163
LOAM-odom	RRE /［（°）.m^-1］	0.0211	0.0191	0.0182	0.0195
LOAM-odom	RTE /%	4.9715	5.9807	3.8240	5.0036
HRegNet-LO-odom	RRE /［（°）.m^-1］	0.0116	0.0099	0.0112	0.0109
HRegNet-LO-odom	RTE /%	4.1401	2.4475	2.4555	3.1177

Tools

Get Citation

Copy Citation Text

Yongjian Fu, Zongchun Li, Hua He, Li Wang, Cong Li. HRegNet-LO: LiDAR Odometry Measurement Based on End-to-End Deep Neural Network[J]. Acta Optica Sinica, 2023, 43(24): 2428003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Remote Sensing and Sensors

Received: Feb. 14, 2023

Accepted: Mar. 22, 2023

Published Online: Dec. 12, 2023

The Author Email: Li Zongchun (13838092876@139.com)

DOI:10.3788/AOS230548

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology