CDAG改进算法及其在GF-6 WFV数据云检测中的应用 下载: 751次
CDAG-Improved Algorithm and Its Application to GF-6 WFV Data Cloud Detection
山东科技大学测绘科学与工程学院, 山东 青岛 266590
图 & 表
图 1. AVIRIS数据(实线)与GF-6 WFV数据(虚线)在部分波段上的光谱响应曲线比较。(a)第一波段范围;(b)第二波段范围;(c)第四波段范围;(d)第六波段范围
Fig. 1. Comparison of AVIRIS(solid line) and GF-6 WFV(dashed line) data spectral response curves at different bands. (a) 1st band; (b) 2nd band; (c) 4th band; (d) 6th band
下载图片 查看原文
图 2. 美国威斯康星州区域的GF-6 WFV原始影像与模拟影像对比。(a)原始影像;(b)模拟影像
Fig. 2. Comparison of raw image and simulated image of the Wisconsin area, USA. (a) Raw GF-6 WFV image; (b) simulated GF-6 WFV image
下载图片 查看原文
图 3. GF-6 WFV数据云检测算法流程图
Fig. 3. Flow chart of GF-6 WFV data cloud detection algorithm
下载图片 查看原文
图 4. 不同种类云覆盖条件下GF-6 WFV数据的云检测结果。(a) 2018-09-02;(b) 2019-06-07;(c) 2019-04-21;(d) 2018-11-03;(e) 2018-12-01;(f) 2018-09-23
Fig. 4. Cloud detection results of GF-6 WFV data under different cloud cover conditions. (a) 2018-09-02; (b) 2019-06-07; (c) 2019-04-21; (d) 2018-11-03; (e) 2018-12-01; (f) 2018-09-23
下载图片 查看原文
图 5. GF-6 WFV数据进行亮地表去除前后的云检测结果。(a) 2018-12-30;(b) 2019-04-21;(c) 2019-05-04
Fig. 5. Cloud detection results of GF-6 WFV data before and after bright surface removal. (a) 2018-12-30; (b) 2019-04-21; (c) 2019-05-04
下载图片 查看原文
图 6. 部分影像矢量化结果。(a)植被地区影像矢量化结果;(b)城镇地区影像矢量化结果
Fig. 6. Partial image vectorization results. (a) Vectorization result of vegetation image; (b) vectorization result of urban area image
下载图片 查看原文
表 1像元库中的典型亮地表像元
Table1. Typical examples of bright surface pixels in the pixel dataset
查看原文
表 2原方法下所选波段组合及阈值
Table2. Band combination and threshold selected under the original CDAG method
Single band | Multi band | Band ratio |
---|
B1>0.27 | B1>0.18 & B4>0.34 | 0.80<B1/B7<1.08 | B7>0.26 | B1>0.22 & B7>0.26 | 0.99<B2/B8<1.20 | | B2>0.14 & B4>0.36 | 0.47<B3/B7<1.14 | | B2>0.18 & B7>0.26 | 0.91<B3/B8<1.07 | | B3>0.14 & B6>0.32 | | | B3>0.14 & B7>0.26 | | | B4>0.36 & B5>0.20 | | | B4>0.36 & B8>0.18 | | | B5>0.20 & B6>0.32 | |
|
查看原文
表 3三波段离差组合所选波段及阈值
Table3. Band and threshold selected by three-band dispersion combination
Band | Average | Multi-band dispersion |
---|
B1,B2,B5 | A1=(B1+B2+B5)/3 | [(B1-A1)/A1+(B2-A1)/A1+(B5-A1)/A1]<0.015 | B1,B2,B5 | A2=(B1+B3+B7)/3 | [(B1-A2)/A2+(B3-A2)/A2+(B7-A2)/A2]<0.023 | B1,B5,B7 | A3=(B1+B5+B7)/3 | [(B1-A3)/A3+(B5-A3)/A3+(B7-A3)/A3]<0.057 | B2,B3,B7 | A4=(B2+B3+B7)/3 | [(B2-A4)/A4+(B3-A5)/A4+(B7-A4)/A4]<0.018 | B2,B5,B7 | A5=(B2+B5+B7)/3 | [(B2-A5)/A5+(B5-A5)/A5+(B7-A5)/A5]<0.055 | B3,B5,B7 | A6=(B3+B5+B7)/3 | [(B3-A6)/A6+(B5-A6)/A6+(B7-A6)/A6]<0.043 |
|
查看原文
表 4亮地表检测算法所选波段及阈值
Table4. Band and threshold selected by the bright surface detection algorithm
Multi band | Multi band+Band difference |
---|
B1>0.30 & B2>0.32 & B7>0.28 | B1>0.31 & B3>0.35 & B3-B7<0.015 | B1>0.31 & B3>0.36 & B7>0.27 | B1>0.31 & B4>0.39 & B4-B7<0.075 | B2>0.33 & B3>0.37 & B7>0.28 | B2>0.34 & B6>0.35 & B6-B7<0.035 | B2>0.34 & B5>0.35 & B7>0.29 | |
|
查看原文
表 5云像元检测结果精度评价
Table5. Accuracy evaluation of cloud pixel detection results
Underlying surface | Image | CR /% | ER /% | MR /% | SR /% |
---|
| 1 | 76.58 | 0.48 | 23.42 | 99.52 | Bare soil | 2 | 88.45 | 9.73 | 11.55 | 91.27 | | 3 | 84.14 | 2.49 | 15.86 | 97.51 | | 1 | 85.46 | 2.79 | 14.54 | 97.21 | Building | 2 | 88.63 | 7.38 | 11.37 | 92.62 | | 3 | 83.79 | 3.42 | 16.21 | 96.58 | | 1 | 90.36 | 0.28 | 9.64 | 99.72 | Vegetation | 2 | 90.87 | 0.66 | 9.13 | 99.34 | | 3 | 88.04 | 0.15 | 11.96 | 99.85 | | 1 | 82.55 | 0.53 | 17.45 | 99.47 | Water | 2 | 77.66 | 0.20 | 22.34 | 99.80 | | 3 | 85.43 | 0.61 | 14.57 | 99.39 | Total | | 85.16 | 2.39 | 14.84 | 97.69 |
|
查看原文
董震, 孙林, 刘喜荣, 王永吉, 梁天辰. CDAG改进算法及其在GF-6 WFV数据云检测中的应用[J]. 光学学报, 2020, 40(16): 1628001. Zhen Dong, Lin Sun, Xirong Liu, Yongji Wang, Tianchen Liang. CDAG-Improved Algorithm and Its Application to GF-6 WFV Data Cloud Detection[J]. Acta Optica Sinica, 2020, 40(16): 1628001.