Remote Sensing Image Cloud and Cloud Shadow Detection Method Based on RDA-Net Model

Chen Zhang; Xiuzai Zhang; Changjun Yang

doi:10.3788/LOP202158.2028005

Laser & Optoelectronics Progress, Volume. 58, Issue 20, 2028005(2021)

Remote Sensing Image Cloud and Cloud Shadow Detection Method Based on RDA-Net Model

Chen Zhang¹, Xiuzai Zhang^1,2、*, and Changjun Yang³

¹School of Electronics and Information, Nanjing University of Information Science & Technology, Nanjing, Jiangsu 210044, China;

²Jiangsu Province Atmospheric Environment and Equipment Technology Collaborative Innovation Center, Nanjing University of Information Science & Technology,Nanjing, Jiangsu 210044, China;

³National Satellite Meteororologistic Center, Key Laboratory of Radiometric Calibration and Validation for Environmental Satellites, China Meteorological Administration, Beijing 100081, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(14)

Fig. 1. Structure of ResBlock and ResUnit. (a) ResBlock; (b) ResUnit

Download full size

Fig. 2. Structure of PAM

Download full size

Fig. 3. Structure of CAM

Download full size

Fig. 4. Structure of R-ASPP

Download full size

Fig. 5. Strecture of RDA-Net

Download full size

Fig. 6. Structure of RU-Net

Download full size

Fig. 7. Experimental dataset. (a) Remote sensing image block; (b) label

Download full size

Fig. 8. Enhanced effects of experimental data. (a) Original image; (b) vertical rotation; (c) horizontal rotation; (d) horizontal and vertical rotation; (e) transformation of brightness; (f) injection of noise; (g) transformation of saturation; (h) transformation of color

Download full size

Fig. 9. Experimental flow of cloud and cloud shadow detection

Download full size

Fig. 10. Relationship between overall accuracy and number of iterations

Download full size

Fig. 11. Comparison of detection results of Gaofen-1 WFV remote sensing image cloud under different methods. (a) Original images; (b) FCN-8s method; (c) K-means method; (d) SegNet method; (e) DeepLab method; (f) RU-Net method; (g) RDA-Net method; (h) cloud tags

Download full size

Fig. 12. Visual comparison of cloud shadow detection results of Gaofen-1 WFV remote sensing image under two methods. (a) Original image; (b) RU-Net method; (c) RDA-Net method; (d) cloud and cloud shadow labels

Download full size

Table 1. Quantitative comparison results of cloud detection by different methods

View table

Table 1. Quantitative comparison results of cloud detection by different methods

Method	P_Precision/%	A_Accuracy /%	R_Recall/%	F₁	M_MIoU
FCN-8s	90.25	84.54	86.38	0.8827	0.7606
K-means	76.42	84.17	72.63	0.7448
SegNet	90.31	93.03	90.72	0.9051	0.7953
DeepLab	92.66	94.86	92.05	0.9235	0.8019
RU-Net	93.80	97.93	92.94	0.9337	0.8375
RDA-Net	94.74	97.82	93.69	0.9421	0.8790

Table 2. Quantitative comparison of cloud shadow detection by different methods
View table
Table 2. Quantitative comparison of cloud shadow detection by different methods
Method P_Precision /% A_Accuracy /% R_Recall /% F₁ M_MIoU
RU-Net 74.67 90.38 68.73 0.7158 0.8375
RDA-Net 85.25 96.04 80.38 0.8274 0.8790

Tools

Get Citation

Copy Citation Text

Chen Zhang, Xiuzai Zhang, Changjun Yang. Remote Sensing Image Cloud and Cloud Shadow Detection Method Based on RDA-Net Model[J]. Laser & Optoelectronics Progress, 2021, 58(20): 2028005

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites