LiDAR Static Mapping Method Based on Spatio‑Temporal Constraints

Mu Zhou; Shaochun Liu; Liangbo Xie; Nan Du

doi:10.3788/AOS241304

Acta Optica Sinica, Volume. 45, Issue 6, 0628005(2025)

LiDAR Static Mapping Method Based on Spatio‑Temporal Constraints

Mu Zhou^1,2、*, Shaochun Liu^1,2, Liangbo Xie^1,2, and Nan Du^1,2,3

¹School of Communications and Information Engineering, Chongqing University of Posts and Telecommunications, Chongqing 400065, China

²Engineering Research Center of Mobile Communications, Ministry of Education, Chongqing 400065, China

³Department of Computer Science and Technology, Tangshan Normal University, Tangshan 063000, Hebei , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(18)

Fig. 1. System block diagram of static mapping method

Download full size

Fig. 2. Schematic diagram of ground height estimation based on height jump

Download full size

Fig. 3. Schematic diagram of scanning frame grouping and test frame selection

Download full size

Fig. 4. Schematic diagram of projected grids

Download full size

Fig. 5. Schematic diagram of dynamic object detection

Download full size

Fig. 6. Recall of ground points and non-ground points in ground candidate points

Download full size

Fig. 7. Ground segmentation performance under different height thresholds. (a) Precision variation; (b) recall variation; (c) $F_{1}$ variation

Download full size

Fig. 8. Performance comparison under different grouping lengths. (a) PR variation; (b) RR variation; (c) $F_{1}^{'}$ variation

Download full size

Fig. 9. Dynamic object detection results under different grouping lengths. (a) Dynamic point benchmark results; (b) grouping length K=3; (c) grouping length K=6; (d) grouping length K=7

Download full size

Fig. 10. Static mapping results of different methods. (a) Original maps; (b) Removert-RM; (c) ERASOR; (d) proposed method

Download full size

Fig. 11. Experimental setup diagram

Download full size

Fig. 12. Dynamic object detection results obtained by different methods in real-world scenarios. (a) Dynamic point benchmark results; (b) Removert-RM; (c) ERASOR; (d) proposed method

Download full size

Fig. 13. Static mapping results obtained by different methods in real-world scenarios. (a) Original map; (b) Removert-RM;

Download full size

Table 1. Key parameters of experiment

View table

Table 1. Key parameters of experiment

Definition	Symbol	Value
Number of grids in radial dimension	$N_{r}$	20
Number of grids in azimuthal dimension	$N_{θ}$	108
Standard deviation threshold	$τ_{v a r}$	0.1
Grid map resolution	$v$	0.3
Dynamic point detection ratio threshold	$ε_{d}$	0.3
Dynamic feature descriptor height threshold	$ε_{g}$	0.1
Test frame interval	$m$	1
Dynamic point inspection depth map resolution	$I$	0.5
Distance ratio threshold	$λ$	0.01
Voxel size	$v_{o x}$	0.5
Search distance	$R$	0.5
Inter-frame matching distance	$τ_{D}$	0.4
Dynamic point clustering inspection threshold	$η$	0.6

Table 2. Comparison of ground segmentation performance of different methods
View table
Table 2. Comparison of ground segmentation performance of different methods
Method Precision Recall F₁
R-GPF^［13］ 66.94 93.34 77.97
Patchwork^［20］ 77.95 90.67 83.83
MapCleaner^［21］ 96.47 79.30 87.05
Proposed 88.10 86.86 87.42

Table 3. Comparison of static mapping qualities of different methods

View table

Table 3. Comparison of static mapping qualities of different methods

Sequence	Method	PR	RR	$F_{1}^{'}$
00	Removert-RM^［11］	85.50	99.35	91.91
	Remover-RM+RV^［11］	86.83	90.62	88.68
	ERASOR^［13］	93.98	97.08	95.50
	Proposed	98.77	95.45	97.08
01	Removert-RM^［11］	94.22	93.60	93.91
	Remover-RM+RV^［11］	95.82	57.08	71.54
	ERASOR^［13］	91.49	95.38	95.38
	Proposed	98.35	91.94	95.04
02	Removert-RM^［11］	76.32	96.80	85.35
	Remover-RM+RV^［11］	83.29	88.37	85.75
	ERASOR^［13］	87.73	97.01	92.14
	Proposed	98.49	96.37	97.42
05	Removert-RM^［11］	86.90	87.88	87.39
	Remover-RM+RV^［11］	88.17	79.98	83.88
	ERASOR^［13］	88.73	98.26	93.25
	Proposed	90.86	97.15	93.90
07	Removert-RM^［11］	80.69	98.82	88.84
	Remover-RM+RV^［11］	82.04	95.50	88.26
	ERASOR^［13］	90.62	99.27	94.75
	Proposed	94.36	95.91	95.13
Average	Removert-RM^［11］	84.73	95.29	89.48
	Remover-RM+RV^［11］	87.23	82.31	84.70
	ERASOR^［13］	90.51	97.40	93.83
	Proposed	96.17	95.36	95.71

Table 4. Comparison of running time for different methods
View table
Table 4. Comparison of running time for different methods
Method Running time /s
Removert-RM^［11］ 0.83
ERASOR^［13］ 0.73
Proposed 1.43

Table 5. Performance comparison in real-world scenarios

View table

Table 5. Performance comparison in real-world scenarios

Method

PR for

avenue /%

RR for

avenue /%

$F_{1}^{'}$ for

avenue /%

Running time

for avenue /s

PR for

parking

garage /%

RR for

parking

garage /%

$F_{1}^{'}$ for

parking

garage /%

Running time for parking garage /s

Removert-RM^［11］

99.46

45.60

62.53

0.72

99.15

43.00

59.99

0.78

ERASOR^［13］

79.70

73.98

76.73

0.58

91.98

53.32

67.51

0.65

Proposed

99.39

86.76

92.65

1.24

89.78

90.28

90.03

1.52

Tools

Get Citation

Copy Citation Text

Mu Zhou, Shaochun Liu, Liangbo Xie, Nan Du. LiDAR Static Mapping Method Based on Spatio‑Temporal Constraints[J]. Acta Optica Sinica, 2025, 45(6): 0628005

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites