Acta Optica Sinica, Volume. 38, Issue 10, 1028002(2018)
Improvement of Universal Dynamic Threshold Cloud Detection Algorithm and Its Application in High Resolution Satellite
Figures & Tables(11)
Fig. 2. Comparison of spectral response function between MODIS and different domestic satellite sensors. (a) The comparison of spectral response functions in red channel; (b) the comparison of spectral response functions in green channel; (c) the comparison of spectral response functions in blue channel; (d) the comparison of spectral response functions in near infrared (NIR) channel
Fig. 3. Sketch map of pixel registration of GF-2 PMS and surface reflectance dataset
Fig. 4. Thick cloud detection results from the improved UDTCDA for ZY3 images. (a) The original false color composite image of ZY3; (b) the corre-sponding cloud detection result of ZY3; (c) the local standard false color composite image of ZY3; (d) the corresponding cloud detection result of local ZY3 image.
Fig. 5. Thin cloud detection results from the improved UDTCDA for GF-1 WFV images. (a) The original false color composite image of ZY3; (b) the corresponding cloud detection result of GF-1 WFV; (c) the local standard false color composite image of GF-1 WFV; (d)the corresponding cloud detection result of local GF-1 WFV image
Fig. 6. Broken cloud detection results from the improved UDTCDA for GF-1 PMS images. (a) The original false color composite image of GF-1 PMS; (b) the corresponding cloud detection result of GF-1 PMS; (c) the local standard false color composite image of GF-1 PMS; (d) the corresponding cloud detection result of local GF-1 PMS image
Table 1. Parameters of dynamic threshold cloud detection model for different satellites
View tableTable 1. Parameters of dynamic threshold cloud detection model for different satellites
Sensor Band a b c GF-1 PMS Blue 0.8125 0.03648 0.1022 Green 0.8015 0.02463 0.07160 Red 0.8456 0.01636 0.04301 NIR 0.8773 0.01062 -0.03471 GF-1 WFV Blue 0.8067 0.03854 0.1248 Green 0.8031 0.02509 0.1248 Red 0.8429 0.01668 0.07558 NIR 0.8742 0.01063 0.03721 GF-2 Blue 0.9404 0.01003 0.1607 Green 0.8096 0.03715 0.07838 Red 0.8036 0.02486 0.05021 NIR 0.8466 0.01654 0.03033 GF-4 Blue 0.8077 0.03673 0.1355 Green 0.7991 0.02390 0.08047 Red 0.8419 0.01666 0.04654 NIR 0.8495 0.01065 0.02873 ZY3 Blue 0.5491 0.2133 0.1090 Green 0.4504 0.1961 0.1090 Red 0.4589 0.1984 0.1054 NIR 0.5009 0.1965 0.05895
Table 2. Spectral conversion parameters between MODIS and multispectral sensors
View tableTable 2. Spectral conversion parameters between MODIS and multispectral sensors
Sensor Band m n R 2GF-1 PMS Blue 1.0514 0.0049 0.9933 Green 0.9866 -0.0008 0.9947 Red 0.9986 0.0088 0.9987 NIR 1.0038 -0.0103 0.9981 GF-1 WFV Blue 1.0907 0.0101 0.9811 Green 1.0005 -0.0024 0.9909 Red 1.0106 0.0110 0.9966 NIR 0.9934 -0.0079 0.9981 GF-2 Blue 0.0098 1.0821 0.9864 Rreen 1.0041 -0.0020 0.9904 Red 0.9662 0.0246 0.9932 NIR 0.9799 -0.0050 0.9981 GF-4 Blue 1.0768 0.0126 0.9873 Green 1.0214 -0.0026 0.9732 Red 1.0146 0.0019 0.9991 NIR 1.0001 -0.0129 0.9968 ZY3 Blue 1.0727 0.0072 0.9880 Green 0.9912 -0.0031 0.9934 Red 1.0011 0.0126 0.9970 NIR 0.9886 -0.0072 0.9980
Table 3. Accuracy verification error matrix of cloud detection results
View tableTable 3. Accuracy verification error matrix of cloud detection results
Item Surface reference information Row total Evaluation index Cloud Clear sky Cloud ρ 11ρ 12ρ 1+ A U=ρ 11/ρ 1+ Detection results Clear sky ρ 21ρ 22ρ 2+ E O=ρ 21/ρ+ 1= 1-A PColumns total ρ+ 1ρ+ 2N - Evaluation index A P=ρ 11/ρ+ 1E C=ρ 12/ρ 1+= 1-A U- A O=(ρ 11+ρ 22)/N
Table 4. Accuracy statistics of cloud detection results
View tableTable 4. Accuracy statistics of cloud detection results
Sensor Number Surface type Cloud type A P/%AU/% E O/%E C/%A O/%ZY3 (a) Vegetation, soil Thick, broken cloud 95.72 95.82 4.28 4.18 96.46 (b) Vegetation Broken, thin cloud 93.81 91.79 6.19 8.21 98.04 (c) Vegetation, soil Thick, broken cloud 92.59 93.49 7.41 6.51 94.70 (d) Desert Thin cloud 92.91 96.38 7.09 3.62 96.06 Average 93.76 94.37 6.24 5.63 96.32 GF-1 WFV (e) Vegetation Thick, thin cloud 83.03 85.89 16.97 14.11 90.84 (f) Soil, vegetation Thick, broken cloud 89.09 91.61 10.91 8.39 93.25 (g) Soil Thick, broken cloud 94.65 97.95 5.35 2.05 95.49 (h) Desert Thin cloud 85.96 90.08 14.04 9.92 96.58 Average 88.18 91.38 11.82 8.62 94.04 GF-1 PMS (i) Vegetation, city Thick, broken cloud 92.09 93.91 7.91 6.09 96.49 (j) Vegetation Thin, thick cloud 85.69 87.06 14.31 12.94 96.14 Average 88.89 90.49 11.11 9.52 96.31 GF-2 (k) Vegetation, soil Thick, thin loud 77.60 79.02 22.40 20.98 83.11 (l) Vegetation Thick cloud 89.87 93.67 10.13 6.33 88.84 Average 83.74 86.35 16.27 13.66 85.98 GF-4 (m) Vegetation Thick, broken cloud 91.65 93.24 8.35 6.76 95.00 (n) Vegetation, city Thin cloud 85.87 88.77 14.13 11.23 92.56 (o) Vegetation, soil Thick, broken cloud 90.51 88.06 9.49 11.94 91.60 (p) Desert Thick, broken cloud 90.39 91.66 9.61 8.34 96.51 Average 89.61 90.43 10.40 9.57 93.92
Tools
Get Citation
Copy Citation Text
Quan Wang, Lin Sun, Jing Wei, Xueying Zhou, Tingting Chen, Meiyan Shu. Improvement of Universal Dynamic Threshold Cloud Detection Algorithm and Its Application in High Resolution Satellite[J]. Acta Optica Sinica, 2018, 38(10): 1028002
Paper Information
Category: Remote Sensing and Sensors
Received: Feb. 12, 2018
Accepted: May. 3, 2018
Published Online: May. 9, 2019
The Author Email:
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20