Airborne Laser Point-Cloud Filtering in Complex Mountainous Terrain Utilizing Deep Global Information Fusion

Jierui Cui; Yunwei Pu; Yan Xia; Yichen Liu

doi:10.3788/LOP240669

Laser & Optoelectronics Progress, Volume. 61, Issue 18, 1815005(2024)

Airborne Laser Point-Cloud Filtering in Complex Mountainous Terrain Utilizing Deep Global Information Fusion

Jierui Cui^1,3, Yunwei Pu^1,2、*, Yan Xia³, and Yichen Liu³

Author Affiliations

¹Faculty of Land and Resources Engineering, Kunming University of Science and Technology, Kunming 650093, Yunnan, China

²Computing Center, Kunming University of Science and Technology, Kunming 650500, Yunnan, China

³Yunnan Water Conservancy and Hydropower Survey and Design Institute, Kunming 650021, Yunnan, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(12)

Fig. 1. Network structure

Download full size

Fig. 2. Local cross feature fusion module

Download full size

Fig. 3. Normal vector difference

Download full size

Fig. 4. Multidimensional feature encoding unit

Download full size

Fig. 5. Global context aggregation module

Download full size

Fig. 6. YNM-dataset

Download full size

Fig. 7. OpenGF dataset

Download full size

Fig. 8. Filtering results of region 3 in the OpenGF dataset. (a) CSF; (b) PMF; (c) PTD; (d) RandLA-Net; (e) SCF-Net; (f) BAAF-Net; (g) MGINet; (h) GT

Download full size

Fig. 9. Filtering results of YNM-dataset. (a) CSF; (b) PMF; (c) PTD; (d) RandLA-Net; (e) SCF-Net; (f) BAAF-Net; (g) MGINet; (h) GT

Download full size

Table 1. Parameter settings for different filtering methods
View table
Table 1. Parameter settings for different filtering methods
Dataset Parameter
PMF CSF PTD
OpenGF dataset WS：1.0，MW：30 CR：0.5，MI：800 MB：60，IA：30，IB：0.5
YNM-dataset WS：0.9，MW：30 CR：0.5，MI：800 MB：40，IA：50，IB：0.5

Table 2. OA, MIoU, and RMSE of different models in two datasets

View table

Table 2. OA, MIoU, and RMSE of different models in two datasets

Method	OpenGF dataset				YNM-dataset
Method	OA	MIoU	RMSE	Time /min	OA	MIoU	RMSE	Time /min
CSF	89.78	84.48	2.21	10.2	86.51	82.61	3.11	45.3
PTD	98.89	93.32	1.84	1.8	92.30	89.65	2.58	7.3
PMF	85.55	80.67	3.93	0.6	76.65	78.58	4.65	3.6
RandLA-Net	97.02	92.29	1.67	1.2	94.32	91.47	1.52	4.5
SCF-Net	96.67	92.49	1.98	1.3	93.26	90.94	2.16	5.6
BAAF-Net	98.37	93.48	1.25	1.1	96.70	93.08	0.97	5.3
MGINet	98.42	94.62	0.89	1.6	96.82	93.29	0.85	6.1

Table 3. Filtering accuracy of each module of MGINet

View table

Table 3. Filtering accuracy of each module of MGINet

MGINet model	OpenGF dataset			YNM-dataset
MGINet model	OA	MIoU	RMSE	OA	MIoU	RMSE
None	93.78	91.10	1.23	92.15	89.35	1.87
Nor-Dist	95.35	92.56	1.06	94.41	91.79	1.25
Cross-feature	94.98	91.54	1.05	93.01	90.26	1.72
GCAM	96.33	93.46	0.90	94.82	92.13	1.06
Nor-Dist+Cross-feature	96.02	93.58	0.98	94.89	92.08	1.18
Nor-Dist+Cross-feature+ GCAM	98.42	94.62	0.89	96.82	93.29	0.85

Tools

Get Citation

Copy Citation Text

Jierui Cui, Yunwei Pu, Yan Xia, Yichen Liu. Airborne Laser Point-Cloud Filtering in Complex Mountainous Terrain Utilizing Deep Global Information Fusion[J]. Laser & Optoelectronics Progress, 2024, 61(18): 1815005

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites