Change Detection of Optical and Synthetic Aperture Radar Remote Sensing Images Based on a Domain Adaptive Neural Network

Qinfeng Yao; Yongxiang Ning; Sunwen Du

doi:10.3788/LOP232565

Laser & Optoelectronics Progress, Volume. 61, Issue 18, 1828001(2024)

Change Detection of Optical and Synthetic Aperture Radar Remote Sensing Images Based on a Domain Adaptive Neural Network

Qinfeng Yao^1、*, Yongxiang Ning¹, and Sunwen Du²

Author Affiliations

¹Department of Earth Science and Engineering, Shanxi Institute of Engineering and Technology, Yangquan 045000, Shanxi, China

²School of Mining Engineering, Taiyuan University of Technology, Taiyuan 030024, Shanxi, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(12)

Fig. 1. Overall architecture of proposed method

Download full size

Fig. 2. Dual attention mechanism of spatial channel. (a) Convolution block attention module; (b) spatial access attention module; (c) spatial attention module; (d) channel attention module

Download full size

Fig. 3. Dataset 1. (a) Landsat-8 optical image; (b) Sentinel-1A SAR image; (c) ground truth

Download full size

Fig. 4. Dataset 2. (a) QuickBird-2 optical image; (b) TerraSAR-X StripMap HH SAR image; (c) ground truth

Download full size

Fig. 5. Dataset 3. (a) Sentinel-2 optical image; (b) COSMO-SkyMed SAR image; (c) ground truth

Download full size

Fig. 6. Change detection results of dataset 1. (a) Optical image; (b) SAR image;(c) DCCN; (d) DHFF; (e) EECD; (f) FCSN; (g) DTCDN; (h) MSCD; (i) proposed method; (j) ground truth

Download full size

Fig. 7. Change detection results of dataset 2. (a) Optical image; (b)SAR image; (c) DCCN; (d) DHFF; (e) EECD; (f) FCSN; (g) DTCDN; (h) MSCD; (i) proposed method; (j) ground truth

Download full size

Fig. 8. Change detection results of dataset 3. (a) Optical images; (b) SAR image;(c) DCCN; (d) DHFF; (e) EECD; (f) FCSN; (g) DTCDN; (h) MSCD; (i) proposed method; (j) ground truth

Download full size

Table 1. Quantitative evaluation results for dataset 1

View table

Table 1. Quantitative evaluation results for dataset 1

Method	$R_{p r e c i s i o n}$	$R_{r e c a l l}$	$R_{m I O U}$	$s_{F 1}$
DCCN	49.92	49.69	38.84	46.31
DHFF	55.75	73.45	44.41	54.50
EECD	73.16	81.92	66.89	76.68
FCSN	77.21	80.02	68.64	78.26
DTCDN	80.29	82.06	71.28	80.67
MSCD	79.89	84.22	72.25	81.54
Proposed method	78.89	85.38	72.74	82.17

Table 2. Quantitative evaluation results for dataset 2

View table

Table 2. Quantitative evaluation results for dataset 2

Method	$R_{p r e c i s i o n}$	$R_{r e c a l l}$	$R_{m I O U}$	$s_{F 1}$
DCCN	47.13	43.91	38.98	45.02
DHFF	52.94	58.17	41.37	51.18
EECD	89.08	91.01	82.35	89.52
FCSN	91.99	94.54	87.17	92.75
DTCDN	88.83	95.17	85.32	91.57
MSCD	91.60	93.31	86.24	92.17
Proposed method	92.04	95.88	88.88	93.86

Table 3. Quantitative evaluation results for dataset 3

View table

Table 3. Quantitative evaluation results for dataset 3

Method	$R_{p r e c i s i o n}$	$R_{r e c a l l}$	$R_{m I O U}$	$s_{F 1}$
DCCN	50.14	50.11	40.57	50.01
DHFF	48.21	47.07	36.53	47.62
EECD	61.6	56.59	48.65	57.71
FCSN	61.59	58.24	48.85	57.74
DTCDN	67.42	65.36	54.92	66.29
MSCD	68.67	74.24	57.43	70.01
Proposed method	71.51	71.91	59.91	71.71

Table 4. Comparison of training time and prediction time
View table
Table 4. Comparison of training time and prediction time
Method Training time Prediction time
EECD 5.98 0.065
FCSN 7.12 0.037
DTCDN 266.15 0.218
MSCD 35.82 0.061
Proposed method 26.19 0.058

Tools

Get Citation

Copy Citation Text

Qinfeng Yao, Yongxiang Ning, Sunwen Du. Change Detection of Optical and Synthetic Aperture Radar Remote Sensing Images Based on a Domain Adaptive Neural Network[J]. Laser & Optoelectronics Progress, 2024, 61(18): 1828001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites