Airborne point cloud classification integrating edge convolution and global-local self-attention

Table 1. Parameter settings
View table
View in Article
Table 1. Parameter settings
K r OA Avg.F1
15 0.1 81.8 65.2
15 0.2 82.1 65.7
15 0.3 81.7 64.9
20 0.1 82.2 65.9
20 0.2 82.5 66.2
20 0.3 82.0 65.7
25 0.1 81.9 65.4
25 0.2 82.3 66.0
25 0.3 82.1 65.8

Table 2. Confusion matrix and evaluation index of the classification results of ISPRS-3D

View table

View in Article

Table 2. Confusion matrix and evaluation index of the classification results of ISPRS-3D

Class	p_l	l_v	i_s	car	f_h	r_f	f_e	s_b	t_e
p_l	33.5	0.3	0.0	0.0	0.0	31.8	11.7	0.2	22.6
l_v	0.0	73.0	9.8	0.2	0.4	1.3	0.4	6.4	8.5
i_s	0.0	3.6	94.9	0.3	0.0	0.3	0.2	0.7	0.1
car	0.0	11.2	14.0	58.2	3.7	2.4	0.6	8.6	1.3
f_h	0.0	12.1	1.6	0.6	24.5	1.2	0.5	48.1	11.4
r_f	0.1	1.4	0.0	0.0	0.0	93.7	1.7	1.4	1.8
f_e	0.2	6.2	0.5	1.6	0.2	23.1	50.1	5.8	12.3
s_b	0.0	15.6	0.4	0.5	1.3	3.2	1.9	53.2	24.0
t_e	0.0	4.7	0.0	0.1	0.2	1.7	1.3	8.7	83.4
Precision	69.6	85.1	88.5	77.6	50.5	92.9	70.9	43.6	77.4
Recall	33.5	73.0	94.9	58.2	24.5	93.7	50.1	53.2	83.4
F1 score	45.9	78.6	91.6	66.5	33.0	93.3	58.7	47.9	80.3

Table 3. Confusion matrix and evaluation index of the classification results of WHU-Urban3D

View table

View in Article

Table 3. Confusion matrix and evaluation index of the classification results of WHU-Urban3D

Class	b_d	t_e	v_g	l_v	l_g	e_c	g_d	o_t	v_h
b_d	93.2	2.7	3.4	0.3	0.1	0.2	0.0	0.2	0.0
t_e	5.2	79.0	9.3	4.7	0.4	0.3	0.0	0.3	0.9
v_g	3.3	2.6	88.5	1.3	0.0	0.0	0.5	2.3	1.6
l_v	4.2	5.2	8.9	74.1	0.0	0.0	0.2	5.1	2.3
l_g	10.6	17.8	8.1	0.2	51.7	7.5	0.0	3.9	0.3
e_c	8.1	4.2	1.3	0.0	6.9	77.0	0.0	2.6	0.0
g_d	0.4	0.1	2.7	1.6	0.2	0.1	93.3	1.3	0.4
o_t	9.8	5.3	7.4	6.0	6.7	8.2	1.9	51.4	3.4
v_h	2.5	1.7	4.3	12.7	1.3	0.0	11.1	7.4	59.0
Precision	92.8	92.2	86.1	74.3	81.3	77.8	92.5	60.8	78.1
Recall	93.2	79.0	88.5	74.1	51.7	77.0	93.3	51.4	59.0
F1 score	93.0	85.1	87.3	74.2	63.2	77.4	92.9	55.7	67.2

Table 4. Ablation experiments on ISPRS-3D dataset

View table

View in Article

Table 4. Ablation experiments on ISPRS-3D dataset

Class	F1 score									OA
Class	p_l	l_v	i_s	car	f_h	r_f	f_e	s_b	t_e	OA
PT	41.4	76.9	89.1	58.8	24.4	91.6	57.9	41.6	77.9	81.1
PT-GA	61.3	79.0	90.0	71.6	25.9	92.6	61.8	36.9	76.9	81.6
PT-EC	51.7	80.5	91.2	58.0	19.9	90.8	60.6	40.1	74.9	81.4
Ours(PT-GA-EC)	45.9	78.6	91.6	66.5	33.0	93.3	58.7	47.9	80.3	82.5

Table 5. Ablation experiments on the WHU-Urban3D dataset

View table

View in Article

Table 5. Ablation experiments on the WHU-Urban3D dataset

Method	F1 score									OA
Method	b_d	t_e	v_g	l_v	l_g	e_c	g_d	o_t	v_h	OA
PT	91.4	79.9	84.7	71.6	52.3	73.8	89.9	44.0	58.2	84.0
PT-GA	92.8	82.5	86.9	75.6	58.3	76.7	93.1	52.3	60.2	85.9
PT-EC	92.2	82.9	86.2	72.9	54.6	75.7	91.3	48.4	68.0	85.5
Ours(PT-GA-E)	93.0	85.1	87.3	74.2	63.2	77.4	92.9	55.7	67.2	87.4

Table 6. Classification results of different methods on ISPRS-3D dataset

View table

View in Article

Table 6. Classification results of different methods on ISPRS-3D dataset

Method	F1 score									OA	Avg.F1
Method	p_l	l_v	i_s	car	f_h	r_f	f_e	s_b	t_e	OA	Avg.F1
PointNet	52.6	70.0	83.3	11.2	7.5	74.8	7.8	24.6	45.4	65.7	41.9
PointNet++	54.7	80.5	90.3	72.4	35.6	89.5	50.2	42.3	73.1	81.0	65.4
DGCNN	45.2	80.6	89.6	77.1	29.6	90.0	55.3	41.2	74.8	81.1	64.9
RandLANet	48.7	78.2	87.3	68.1	37.2	91.1	57.8	45.3	77.4	81.5	65.6
PointMLP	54.8	78.3	90.8	69.8	32.2	89.4	56.1	41.5	76.3	81.4	65.4
Stratified transformer	51.8	80.9	90.5	66.7	35.1	90.2	56.4	44.2	76.0	81.6	65.7
Ours	45.9	78.6	91.6	66.5	33.0	93.3	58.7	47.9	80.3	82.5	66.2

Table 7. Classification results of different methods on the WHU-Urban3D dataset unit

View table

View in Article

Table 7. Classification results of different methods on the WHU-Urban3D dataset unit

Method	F1 score									OA	Avg.F1
Method	b_d	t_e	v_g	l_v	l_g	e_c	g_d	o_t	v_h	OA	Avg.F1
PointNet	81.6	54.3	73.8	43.5	51.7	67.8	88.5	31.3	46.2	70.5	59.8
PointNet++	90.1	75.3	83.5	71.3	62.7	82.5	91.3	46.4	71.1	85.3	74.9
DGCNN	91.3	77.8	87.4	70.8	60.1	74.5	91.7	47.2	77.4	85.6	75.3
RandLANet	91.7	79.3	88.1	71.8	63.5	79.7	91.5	53.7	66.9	86.2	76.2
PointMLP	92.4	80.8	86.2	73.1	62.7	75.3	90.3	54.5	68.2	86.0	75.9
Stratified transformer	92.6	81.2	87.5	72.6	64.1	76.5	91.2	53.8	68.5	86.5	76.4
Ours	93.0	85.1	87.3	74.2	63.2	77.4	92.9	55.7	67.2	87.4	77.3

Table 8. Classification results of the proposed method and the existing methods on the ISPRS 3D

View table

View in Article

Table 8. Classification results of the proposed method and the existing methods on the ISPRS 3D

Method	F1 score									OA	Avg.F1
Method	p_l	l_v	i_s	car	f_h	r_f	f_e	s_b	t_e	OA	Avg.F1
UM	46.1	79.0	89.1	47.7	5.2	92.2	52.7	40.9	77.9	80.8	59.0
WhuY2	31.9	80.0	88.9	40.8	24.5	93.1	49.4	41.1	77.3	81.0	58.6
WhuY3	37.1	81.4	90.1	63.4	23.9	93.4	47.5	39.9	78.0	82.3	61.6
BIJ_W	13.8	78.5	90.5	56.4	36.3	92.2	53.2	43.3	78.4	81.5	60.3
RIT_l	37.5	77.9	91.5	73.4	18.0	94.0	49.3	45.9	82.5	81.6	63.3
Ours	45.9	78.6	91.6	66.5	33.0	93.3	58.7	47.9	80.3	82.5	66.2

Tools

Get Citation

Copy Citation Text

Jingmin TU, Jin YAN, Li LI, Jian YAO, Jie LI, Yanfei KANG. Airborne point cloud classification integrating edge convolution and global-local self-attention[J]. Optics and Precision Engineering, 2024, 32(24): 3658

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category:

Received: Jul. 8, 2024

Accepted: --

Published Online: Mar. 11, 2025

The Author Email: LI Jie (jielonline@hbut.edu.cn)

DOI:10.37188/OPE.20243224.3658

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Parameter settings

Table 1. Parameter settings

Table 2. Confusion matrix and evaluation index of the classification results of ISPRS-3D

Table 2. Confusion matrix and evaluation index of the classification results of ISPRS-3D

Table 3. Confusion matrix and evaluation index of the classification results of WHU-Urban3D

Table 3. Confusion matrix and evaluation index of the classification results of WHU-Urban3D

Table 4. Ablation experiments on ISPRS-3D dataset

Table 4. Ablation experiments on ISPRS-3D dataset

Table 5. Ablation experiments on the WHU-Urban3D dataset

Table 5. Ablation experiments on the WHU-Urban3D dataset

Table 6. Classification results of different methods on ISPRS-3D dataset

Table 6. Classification results of different methods on ISPRS-3D dataset

Table 7. Classification results of different methods on the WHU-Urban3D dataset unit

Table 7. Classification results of different methods on the WHU-Urban3D dataset unit

Table 8. Classification results of the proposed method and the existing methods on the ISPRS 3D

Table 8. Classification results of the proposed method and the existing methods on the ISPRS 3D