基于光谱分析和动态分形维数的高光谱遥感图像云检测 
下载： 1084次
Hyperspectral Remote Sensing Image Cloud Detection Based on Spectral Analysis and Dynamic Fractal Dimension
1 浙江大学电气工程学院, 浙江 杭州 310027
2 杭州电子科技大学计算机应用技术研究所, 浙江 杭州 310018
3 上海卫星工程研究所, 上海 200240
图 & 表
图 1. 含厚云、薄云、雪、海洋、陆地的遥感图像及其光谱曲线。(a)遥感图像;(b)光谱曲线
Fig. 1. Remote sensing image with thick clouds, thin clouds, snow, sea, and land and their spectral curves. (a) Remote sensing image; (b) spectral curves
下载图片 查看原文
图 2. 基于动态分形维数和辐射量特性的云检测算法流程图
Fig. 2. Flow chart of cloud detection algorithm based on dynamic fractal dimension and radiation characteristics
下载图片 查看原文
图 3. 遥感云检测样本图像。(a)样本A;(b)样本B;(c)样本C;(d)样本D;(e)样本E;(f)样本F;(g)样本G;(h)样本H
Fig. 3. Sample images of remote sensing cloud detection. (a) Sample A; (b) sample B; (c) sample C; (d) sample D; (e) sample E; (f) sample F; (g) sample G; (h) sample H
下载图片 查看原文
图 4. 真值云图。(a)样本A;(b)样本B;(c)样本C;(d)样本D;(e)样本E;(f)样本F;(g)样本G;(h)样本H
Fig. 4. True value cloud maps. (a) Sample A; (b) sample B; (c) sample C; (d) sample D; (e) sample E; (f) sample F; (g) sample G; (h) sample H
下载图片 查看原文
图 5. MRC算法检测结果。(a)样本A;(b)样本B;(c)样本C;(d)样本D;(e)样本E;(f)样本F;(g)样本G;(h)样本H
Fig. 5. Detection results of MRC algorithm. (a) Sample A; (b) sample B; (c) sample C; (d) sample D; (e) sample E; (f) sample F; (g) sample G; (h) sample H
下载图片 查看原文
图 6. DFD_RC算法检测结果。(a)样本A;(b)样本B;(c)样本C;(d)样本D;(e)样本E;(f)样本F;(g)样本G;(h)样本H
Fig. 6. Detection results of DFD_RC algorithm. (a) Sample A; (b) sample B; (c) sample C; (d) sample D; (e) sample E; (f) sample F; (g) sample G; (h) sample H
下载图片 查看原文
图 7. MRC算法的部分放大图像。(a)样本B;(b)样本H;(c)样本G;(d)样本C
Fig. 7. Partially enlarged images of MRC algorithm. (a) Sample B; (b) sample H; (c) sample G; (d) sample C
下载图片 查看原文
图 8. DFD_RC算法的部分放大图像。(a)样本B;(b)样本H;(c)样本G;(d)样本C
Fig. 8. Partially enlarged images of DFD_RC algorithm. (a) Sample B; (b) sample H; (c) sample G; (d) sample C
下载图片 查看原文
表 1遥感云检测样本图像信息
Table1. Information of remote sensing cloud detection sample images
| Sample number | Imaging time | Location | Type of land | Type of cloud |
|---|
| Sample A | 2017-02-03 | Hong Kong | Sea | Thick cloud (small quantity) | | Sample B | 2016-09-06 | Japan Island | Sea | Thick cloud (medium quantity) | | Sample C | 2017-04-23 | Caribbean Sea | Sea | Thick cloud (large quantity) | | Sample D | 2017-02-03 | Hong Kong | Sea, Land | Thin cloud (small quantity) | | Sample E | 2017-01-17 | Norway | Sea, Land, Snow | Thick and thin cloud (small quantity) | | Sample F | 2017-04-23 | Caribbean Sea | Sea, land | Thick and thin cloud (small quantity) | | Sample G | 2017-11-30 | New York | Land | Thin cloud (large quantity) | | Sample H | 2017-11-30 | New York | Land | Thick cloud (small quantity) |
|
查看原文
表 2真值云图的云含量占比
Table2. Cloud content ratios of true value cloud maps %
| Parameter | Sample A | Sample B | Sample C | Sample D | Sample E | Sample F | Sample G | Sample H |
|---|
| Cloud ratio | 11.6415 | 30.9074 | 69.3540 | 37.0145 | 13.4696 | 10.0031 | 43.5738 | 7.0778 | | Thick cloud ratio | 7.8491 | 22.6727 | 50.5277 | 7.3675 | 7.2262 | 7.1581 | 4.7550 | 5.9434 | | Thin cloud ratio | 3.7924 | 8.2347 | 18.8264 | 29.6471 | 6.2434 | 2.8449 | 38.8188 | 1.1343 |
|
查看原文
表 3云含量检测结果
Table3. Detection results of cloud content%
| Parameter | Sample A | Sample B | Sample C | Sample D | Sample E | Sample F | Sample G | Sample H |
|---|
| Cloud ratio byMRC | 12.2374 | 24.1508 | 28.4241 | 1.8609 | 20.8452 | 7.3786 | 2.6832 | 0.6023 | | Thick cloud ratioby MRC | 5.8172 | 10.4733 | 1.0546 | 0.3762 | 9.3589 | 6.1693 | 2.6832 | 0.6023 | | Thin cloud ratioby MRC | 6.4202 | 13.6775 | 27.3695 | 1.4847 | 11.4863 | 1.2093 | 0 | 0 | | Cloud ratio byDFD_RC | 11.5851 | 28.2615 | 64.7071 | 13.0741 | 13.0026 | 9.7690 | 38.4335 | 6.7717 | | Thick cloud ratioby DFD_RC | 7.9308 | 19.5402 | 52.0699 | 7.8912 | 7.5631 | 9.0374 | 7.9997 | 6.1051 | | Thin cloud ratioby DFD_RC | 3.6542 | 8.7213 | 12.6372 | 5.1829 | 5.4395 | 0.7316 | 30.4338 | 0.6667 |
|
查看原文
表 4分类指标
Table4. Classification indexes
| Parameter | Sample A | Sample B | Sample C | Sample D | Sample E | Sample F | Sample G | Sample H |
|---|
| Recall rate by MRC /% | 95.5 | 90.3 | 54.7 | 8.5 | 86.4 | 88.1 | 8.7 | 10.5 | | Leak alarm by MRC /% | 4.5 | 9.7 | 45.3 | 91.5 | 13.6 | 11.9 | 91.3 | 89.5 | | Accuracy rate by MRC /% | 97.2 | 96.5 | 99.6 | 97.8 | 92.4 | 96.3 | 99.1 | 98.7 | | False alarm by MRC /% | 2.8 | 3.5 | 0.4 | 2.2 | 7.6 | 3.7 | 0.9 | 1.3 | | Recall rate by DFD_RC /% | 97.7 | 97.2 | 98.6 | 74.3 | 93.8 | 95.5 | 90.0 | 91.1 | | Leak alarm by DFD_RC /% | 2.3 | 2.8 | 1.4 | 25.7 | 6.2 | 4.5 | 10.0 | 8.9 | | Accuracy rate by DFD_RC /% | 98.8 | 96.3 | 99.6 | 98.7 | 87.3 | 96.1 | 96.7 | 96.5 | | False alarm by DFD_RC /% | 1.2 | 3.7 | 0.4 | 1.3 | 12.7 | 3.9 | 3.3 | 3.5 | | Mean time by MRC /s | 0.3807 | | Mean time by DFD_RC /s | 0.7900 |
|
查看原文
徐冬宇, 厉小润, 赵辽英, 舒锐, 唐琪佳. 基于光谱分析和动态分形维数的高光谱遥感图像云检测[J]. 激光与光电子学进展, 2019, 56(10): 101003. Dongyu Xu, Xiaorun Li, Liaoying Zhao, Rui Shu, Qijia Tang. Hyperspectral Remote Sensing Image Cloud Detection Based on Spectral Analysis and Dynamic Fractal Dimension[J]. Laser & Optoelectronics Progress, 2019, 56(10): 101003.
PDF全文
