Improved ResUNet Method for Extracting Buildings from Remote Sensing Images

Fig. 8. Building extraction results on WHU dataset. (a) Original images; (b) labels; (c) ResUNet; (d) HRNetV2; (e) MSFCN; (f) BuildFormer; (g) DC-Swin; (h) SDSC-UNet; (i) SGMFResUNet

Download full size

Fig. 9. Building extraction results on Massachusetts dataset. (a) Original images; (b) labels; (c) ResUNet; (d) HRNetV2; (e) MSFCN; (f) BuildFormer; (g) DC-Swin; (h) SDSC-UNet; (i) SGMFResUNet

Download full size

Table 1. Comparison of extraction accuracy of different networks on WHU dataset

View table

Table 1. Comparison of extraction accuracy of different networks on WHU dataset

Network	$R_{I o U}$ /%	$R_{M I o U}$ /%	$R_{P r e c i s i o n}$ /%	$R_{R e c a l l}$ /%	$s_{F 1}$ /%	Parameter /Mbit	Test time per picture /s
ResUNet	88.81	93.30	92.90	95.27	94.07	9.601	0.066
HRNetV2	90.36	94.23	93.69	96.03	94.85	28.545	0.091
MSFCN	90.58	94.39	95.75	94.43	95.09	14.170	0.069
BuildFormer	90.29	94.17	95.21	94.35	94.78	38.354	0.064
DC-Swin	87.37	92.46	94.51	92.04	93.26	66.901	0.073
SDSC-UNet	90.46	94.31	95.60	94.39	94.99	21.319	0.085
SGMFResUNet	90.74	94.46	94.23	96.08	95.15	24.721	0.102

Table 2. Comparison of extraction accuracy of different networks on Massachusetts dataset

View table

Table 2. Comparison of extraction accuracy of different networks on Massachusetts dataset

Network	$R_{I o U}$ /%	$R_{M I o U}$ /%	$R_{P r e c i s i o n}$ /%	$R_{R e c a l l}$ /%	$s_{F 1}$ /%	Parameters /M	Test time per picture/s
ResUNet	73.71	84.22	80.83	89.34	84.87	9.601	0.067
HRNetV2	75.16	85.12	82.07	89.93	85.82	28.545	0.088
MSFCN	75.48	85.42	85.42	86.64	86.02	14.170	0.069
BuildFormer	75.65	85.51	85.49	86.79	86.14	38.354	0.062
DC-Swin	71.20	82.82	82.46	83.91	83.18	66.901	0.073
SDSC-UNet	76.16	85.85	86.27	86.65	86.46	21.319	0.084
SGMFResUNet	77.00	86.24	83.12	91.29	87.01	24.721	0.103

Table 3. Comparison of ablation experimental indexes

View table

Table 3. Comparison of ablation experimental indexes

Dataset	Network	$R_{I o U}$	$R_{M I o U}$	$R_{P r e c i s i o n}$	$R_{R e c a l l}$	$s_{F 1}$
WHU	Baseline	89.62	93.78	93.49	95.58	94.52
	Baseline+SIE-GAM	89.80	93.89	93.54	95.74	94.63
	Baseline+HDEM	89.95	93.91	93.67	95.90	94.77
	Baseline+DPDP	90.13	94.09	93.90	95.73	94.81
	Baseline+DPDP+ACB	90.04	94.04	93.98	95.55	94.76
	Baseline+DPDP+HDEM	90.12	94.08	93.69	95.95	94.80
	Baseline+DPDP+ACB+HDEM	90.37	94.24	94.21	95.66	94.93
	Baseline+DPDP+ACB+HDEM+SIE-GAM （SGMFResUNet）	90.74	94.46	94.23	96.08	95.15
Massachusetts	Baseline	73.52	84.06	79.63	90.55	84.74
	Baseline+SIE-GAM	75.42	85.37	82.05	90.94	86.27
	Baseline+HDEM	74.38	84.62	80.94	90.17	85.31
	Baseline+DPDP	73.66	84.18	80.46	89.71	84.83
	Baseline+DPDP+ACB	74.76	84.85	81.26	90.33	85.56
	Baseline+DPDP+HDEM	73.90	84.30	80.21	90.38	84.99
	Baseline+DPDP+ACB+HDEM	75.39	85.24	81.63	90.80	85.97
	Baseline+DPDP+ACB+HDEM+SIE-GAM （SGMFResUNet）	77.00	86.24	83.12	91.29	87.01

Tools

Get Citation

Copy Citation Text

Qianrong Sun, Xiaopeng Wang. Improved ResUNet Method for Extracting Buildings from Remote Sensing Images[J]. Laser & Optoelectronics Progress, 2025, 62(8): 0828003

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: Sep. 4, 2024

Accepted: Oct. 28, 2024

Published Online: Apr. 8, 2025

The Author Email:

DOI:10.3788/LOP241955

CSTR:32186.14.LOP241955

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology