Please wait a minute...
img

官方微信

遥感技术与应用  2020, Vol. 35 Issue (6): 1337-1347    DOI: 10.11873/j.issn.1004-0323.2020.6.1337
冰雪遥感专栏     
基于FY-4A/AGRI时空特征融合的新疆地区积雪判识
张永宏1,3(),曹海啸1,阚希2()
1.南京信息工程大学 自动化学院,江苏 南京 210044
2.南京信息工程大学 滨江学院,江苏 无锡 214105
3.南京信息工程大学江苏省大气环境与装备技术协同创新中心,江苏 南京 210044
Snow Cover Recognition for Xinjiang based on Fusion of FY-4A/AGRI Spatial and Temporal Characteristics
Yonghong Zhang1,3(),Haixiao Cao1,Xi Kan2()
1.College of Automation,Nanjing University of Information Science& Technology,Nanjing 210044,China
2.Bingjiang College,Nanjing University of Information Science& Technology,Wuxi 214105,China
3.Jiangsu Collaborative Innovation Center of Atmospheric Environment and Equipment Technology(CICAEET),Nanjing University of Information Science &Technology,Nanjing 210044,China
 全文: PDF(15352 KB)   HTML
摘要:

高时间分辨率的积雪判识对于新疆牧区农牧业发展和雪灾预警具有重要作用,针对已有积雪产品易受复杂地形地貌,下垫面类型以及云遮蔽的影响,导致积雪判识精度降低的问题,提出一种利用深度学习方法对风云4号A星多通道辐射扫描计(AGRI)数据与地理信息数据进行多特征时序融合的积雪判识方法:以多时相FY-4A/AGRI多光谱遥感数据,以及高程、坡向、坡度和地表覆盖类型等地形地貌信息作为模型输入,以Landsat 8 OLI提取的高空间分辨率积雪覆盖图作为“真值”标签,构建并训练基于卷积神经网络的积雪判识模型,从而有效区分新疆复杂地形与下垫面地区的云、雪以及无雪地表,最终得到逐小时积雪覆盖范围产品。经数据集和2019年地面气象站实测雪盖验证,该方法精度高于国际主流MODIS逐日积雪产品MOD10A1和MYD10A1,显著降低云雪误判率。

关键词: 新疆深度学习积雪FY?4A/AGRIMOD10A1    
Abstract:

Snow cover recognition with high temporal resolution plays an important role in the development of agriculture and animal husbandry and snow disaster warning in Xinjiang pastoral areas. To solve the problem that existing snow cover products are susceptible to complex topography, landform, underlying surface type and cloud cover, which leads to the reduced accuracy of snow cover recognition, a deep learning method is proposed to use the data of Fengyun-4A Star Multichannel Radiation Scanner (AGRI) and the number of geographic information.Based on the method of multi-feature time series fusion, a new snow cover recognition model based on convolution neural network is constructed and trained, which takes the multitemporal FY-4A/AGRI multispectral remote sensing data, terrain topographic information such as elevation, aspect, slope, and surface cover type as the input of the model, and the high-resolution snow cover map extracted by Landsat 8-OLI as the "true value" label.Clouds, snow and snow-free surfaces in Xinjiang's complex terrain and underlying areas ultimately lead to hourly snow cover products. It is verified by the data set and the snow coverof meteorological station in 2019 the accuracy of this method is higher than that of MOD10A1 and MYD10A1, the main international MODIS snow products, which significantly reduces the misclassification rate of cloud and snow.

Key words: Xinjiang    Deep learning    Snow cover    Fengyun-4A/AGRI    MOD10A1
收稿日期: 2020-07-31 出版日期: 2021-01-26
ZTFLH:  P237  
基金资助: 国家自然科学基金面上项目“基于天气系统自动识别的新疆牧区雪灾遥感监测与预警研究”(41875027)
通讯作者: 阚希     E-mail: zyh@nuist.edu.cn;kanxi@nuist.edu.cn
作者简介: 张永宏(1974—),男,山东临沂人,教授,主要从事大气遥感检测、图像处理分析研究。E?mail:zyh@nuist.edu.cn
服务  
把本文推荐给朋友
加入引用管理器
E-mail Alert
RSS
作者相关文章  
张永宏
曹海啸
阚希

引用本文:

张永宏,曹海啸,阚希. 基于FY-4A/AGRI时空特征融合的新疆地区积雪判识[J]. 遥感技术与应用, 2020, 35(6): 1337-1347.

Yonghong Zhang,Haixiao Cao,Xi Kan. Snow Cover Recognition for Xinjiang based on Fusion of FY-4A/AGRI Spatial and Temporal Characteristics. Remote Sensing Technology and Application, 2020, 35(6): 1337-1347.

链接本文:

http://www.rsta.ac.cn/CN/10.11873/j.issn.1004-0323.2020.6.1337        http://www.rsta.ac.cn/CN/Y2020/V35/I6/1337

图1  技术流程图
图2  新疆地理高程和标签数据来源
图3  AGRI遥感数据预处理流程
通道序号中心波长/μm位深通道分辨率
10.471210 992×10 992
20.651221 984×21 984
30.831210 992×10 992
41.37125 496×5 496
51.61125 496×5 496
62.22125 496×5 496
73.72125 496×5 496
表1  AGRI 2 km 通道信息
图4  2018年10月2日轨道行列号为035146的Landsat 8 OLI影像积雪提取方式
图5  二维卷积网络结构图
图6  深度网络积雪判识模型
操作步骤网络层名称参数输出特征大小
空间 特征 提取输入层\7×7×13
第一次卷积卷积核=3×3×16,步长=15×5×16
第二次卷积卷积核=3×3×32,步长=13×3×32
5时相输入1 440×1
时序 特征 提取卷积层C1卷积核=3×1×16,步长=11 438×16
激活函数Rule\
池化层P1窗口=3×1,步长=3479×16
卷积层C2卷积核=3×16×32,步长=1477×32
激活函数Rule\
池化层P2窗口=3×1,步长=3159×32
卷积层C3卷积核=3×32×64,步长=1157×64
激活函数Rule\
全连接层权值=10 048×1010
Softmax3
表2  网络结构参数
行列号过境时间行列号过境时间
03114813点27分03314813点28分
03214813点28分03414813点29分
03315013点41分03414713点22分
03314913分34分03514713点23分
表3  OLI过境时间
实验

截取的像

素块大小

空间特征提取网络/(卷积核数)时序特征提取网络/(卷积核数)模型输入数据
一层二层三层一层二层三层四层池化步长
A7×716//16///3AGRI 7通道+地理信息数据
B7×71632/1632//3AGRI 7通道+地理信息数据
C7×71632/163264/3AGRI 7通道+地理信息数据
D7×71632/1632641283AGRI 7通道+地理信息数据
E7×71632641632641283AGRI 7通道+地理信息数据
F7×71632/163264/2AGRI 7通道+地理信息数据
G7×71632/163264/4AGRI 7通道+地理信息数据
H5×51632/163264/3AGRI 7通道+地理信息数据
I9×91632/163264/3AGRI 7通道+地理信息数据
J7×71632/163264/3AGRI 7通道
K7×71632/163264/3AGRI 7通道+NDSI
L7×71632/163264/3AGRI 7通道+NDVI
M7×71632/163264/3AGRI 7通道+DEM
N7×71632/163264/3AGRI 7通道+坡向
O7×71632/163264/3AGRI 7通道+坡度
P7×71632/163264/3AGRI 7通道+地表覆盖类型
表4  网络结构参数
实验测试集准确度训练集准确度
裸地综合裸地综合
MOD10A10.931 00.940 30.949 60.940 3////
MYD10A10.929 80.941 30.948 30.939 8////
A0.746 80.669 70.769 80.728 80.766 90.689 50.789 80.748 9
B0.905 60.897 40.918 60.901 20.924 50.914 50.938 60.924 3
C0.941 90.936 90.959 60.946 80.954 50.956 60.974 60.961 9
D0.908 90.927 80.924 50.920 80.918 70.924 50.934 80.926 4
E0.919 70.921 50.929 80.929 40.910 90.917 80.931 00.929 1
F0.895 40.896 70.901 90.898 60.901 80.908 70.908 90.908 6
G0.895 40.896 70.901 90.898 60.901 80.908 70.908 90.908 6
H0.935 80.938 80.949 80.935 60.947 80.948 90.939 80.945 6
I0.935 60.931 20.941 20.936 40.945 80.945 50.938 90.943 8
J0.925 40.928 70.946 70.933 60.935 40.937 80.963 60.945 6
K0.927 70.928 90.946 90.934 50.942 10.948 20.968 40.952 9
L0.933 20.928 70.939 80.933 90.957 90.934 60.956 30.949 6
M0.933 20.936 90.953 50.941 20.936 40.948 60.968 90.951 3
N0.930 90.931 20.942 30.934 80.950 70.946 30.963 80.953 6
O0.933 90.937 20.942 30.937 80.946 90.938 50.961 30.948 9
P0.934 30.934 60.946 30.938 40.948 00.946 30.966 50.953 6
表5  多方法实验结果
数据编号过境时间轨道行号轨道列号积雪分类精度总精度
M12018/09/250351450.938 90.958 7
M22018/10/020351460.934 60.938 9
M32018/11/240341490.956 30.954 6
M42019/11/190301410.926 40.934 6
M52019/12/050291410.926 80.936 4
M62020/02/070291410.945 80.948 2
表6  验证集Landsat8-OLI相关信息
图7  OLI彩色图像与本文方法积雪提取对比
图8  新疆2018年12月6日13点45分多种方法积雪提取效果对比 审图号:GS(2016)3333号
图9  多种方法积雪提取局部效果对比
图10  新疆2019年积雪期周积雪精度对比
周号积雪分类精度/(%)
MYD10A1MOD10A1本文方法
1~9、51、5293.7693.1994.15
45~5070.9669.5671.36
表7  新疆2019年积雪期积雪精度对比
1 Dahe Qing,Zhou Botao,Xiao Cunde. Changes of the Cryosphere and Its Impact on Climate in China[J]. Journal of Meteorology, 2014,72(5):869-879.
1 秦大河,周波涛,效存德.冰冻圈变化及其对中国气候的影响[J].气象学报,2014,72(5):869-879.
2 Xia K, Luo Y, Li W P.Simulation of Freezing and Melting of Soil on the Northeast Tibetan Plateau[J].Chinese Science Bulletin,2011,56(20):2145-2155.
3 Yu Hongfeng,Zhang Xianfeng. Analysis of Snow Cover Change in Xinjiang based on Optical and Microwave Remote Sensing[J]. Earth Information Science, 2015,17(2):244-252.
3 于泓峰,张显峰.光学与微波遥感的新疆积雪覆盖变化分析[J].地球信息科学,2015,17(2):244-252.
4 Yan Xueying, Zhang Qinqin, Zhang Sicong, et al. Monitoring of Glacier and Perennial Snow Cover in Xinjiang from 2005 to 2015[J]. Geospatial Information, 2019,17(8):36-39,10.
4 鄢雪英,张琴琴,张思聪,等.2005~2015年新疆冰川与常年积雪变化监测研究[J].地理空间信息,2019,17(8):36-39,10.
5 Ma Yu, Wang Yun, Wang Xu. Classification of White Disasters and Circulation Characteristics of Disaster-induced Snowstorm in Altai Area[J]. Resources and Environment in Arid Areas,2014,28(8):120-124.
5 马禹,王昀,王旭.阿勒泰地区白灾的等级划分及致灾暴雪的环流特征[J].干旱区资源与环境,2014,28(8):120-124.
6 Wang Xiaoru, Tang Zhiguang, Wang Jian, et al. Remote Sensing Monitoring of Snow Line Height at the End of Snowmelt in High Asia based on MODIS Snow Products[J]. Journal of Geography, 2020,75(3): 470-484.
6 王晓茹,唐志光,王建,等.基于MODIS积雪产品的高亚洲融雪末期雪线高度遥感监测[J].地理学报,2020,75(3):470-484.
7 Pei H, Fang S F, Liu Z H,et al. Snow Information Abstraction based on Remote Sensing Data: Taking the North of Xinjiang for Example[J]. Taylor and amp; Francis Group,2012,12(1):56-60.
8 Dai L Y, Che T, Wang J,et al. Snow Depth and Snow Water Equivalent Estimation from AMSR-E Data based on a Priori Snow Characteristics in Xinjiang,China[J].Remote Sensing of Environment,2012,127(1):14-29.
9 Liu Y, Cheng X,Hao J S,et al. Snow Cover Estimation from MODIS and Sentinel-1 SAR Data Using Machine Learning Algorithms in the Western Part of the Tianshan Mountains[J].Journal of Mounta in Science,2020,17(4):884-897.
10 Sun Yuyan,Zhang Lei, Lu Shanlong,et al. Daily Snow Monitoring Method based on Dynamic NDSI Threshold[J]. Journal of Earth Information Science, 2020,22(2): 298-307.
10 孙玉燕,张磊,卢善龙,等.基于动态NDSI阈值的每日积雪监测方法[J].地球信息科学学报,2020,22(2):298-307.
11 Liu Y,Chen X,Hao J S,et al.Snow Cover Estimation from MODIS and Sentinel-1 SAR Data Using Machine Learning Algorithms in the Western Part of the Tianshan Mountains[J]. Journal of Mountain Science, 2020,17(4):884-897.
12 Hall D K, Riggs G A. In: National Snow and Ice Data Center (NSIDC) (Ed.), MODIS/Terra Snow Cover Daily L3 Global 500 m Grid, Version 6. NASA, Boulder, Colorado USA[J]. Quaternary Science Reviews, 2016,32(1):6-15.
13 Hall D K,Riggs G A,Salomonson V V, et al. MODIS Snow Cover Products[J]. Remote Sensing of Environment, 2002,83(1):181-194.
14 Salomonson V,Appel I.Estimation Fractional Snow Cover from MODIS Using the Normalized Difference Snow Index[J].Remote Sensing of Environment,2004,89(3):351-360.
15 Wang Xuan,Hao Xiaohua,Wang Jian,et al.Accuracy Evaluation of Long Time Series AVHRR Snow Cover Area Products in China[J].Remote Sensing Technology and Application, 2018,33(6):994-1003.
15 王轩,郝晓华,王建,等.中国地区AVHRR长时间序列积雪范围产品精度评估[J].遥感技术与应用,2018,33(6):994-1003.
16 Yan J N, Ma Y, Wang L Z,et al.A Cloud-based Remote Sensing Data Production System[J].Future Generation Computer Systems,2018,86(3):1154-1166.
17 Feng Xun,Li Libing,Cheng Boyang,et al. On Orbit Calibration and Performance Evaluation of FY-4A Atmospheric Vertical Detector[J]. Journal of Infrared and Millimeter Wave, 2019,38(5):648-654.
17 冯绚,李利兵,陈博洋,等.风云四号A星干涉式大气垂直探测仪在轨定标及性能评价[J].红外与毫米波学报,2019,38(5):648-654.
18 Dong Yunya,Zhang Qian.A Survey of Depth Semantic Feature Extraction of High-resolution Remote Sensing Images based on CNN[J].Remote Sensing Technology and Application,2019,34(1):1-11.
18 董蕴雅,张倩.基于CNN的高分遥感影像深度语义特征提取研究综述[J].遥感技术与应用,2019,34(1):1-11.
19 Krizhevsky A,Sutskever I,Hinton G E. ImageNet Classification with Deep Convolutional Neural Networks[C]∥ Advances in Neural Information Processing Systems.Red Hook,NY:Curran Associate,2012:1097-1105.
20 Mikolov T,Karafiat M,Burget L,et al. Recurrent Nerual Network based Language Model[C]∥ Proceedings of INTE-RSPEECH.Lyon France:ISCA,2010:1045-1048.
21 Hochreiter S,Schmidhuber J.Long Short-term Memory.[J].Neural Computation,1997,9(8):1735-1780.
22 Hinton G,Osindero S,Teh Y W,et al.Learning Algorithm for Deep Belief Nets[J].Neural Computation,2006,18(7):1527-1554.
23 Vincent P, Larochelle H, Bengio Y,et al.Extracting and Composing Robust Features with Denoising Autoencoders[C]∥ Proceedings of the 25th International Conference on Machine Learning.New York:ACM Press,2008:1096-1103.
24 Zhang Jiahua,Wu Yang, Yao Fengmei,et al.Characteristics of Snow Cover in Xinjiang in Recent Years based on Satellite Remote Sensing and Ground Measured Snow Data[J].Plateau Meteorology, 2008,27(3):551-557.
24 张佳华,吴杨,姚凤梅,等.利用卫星遥感和地面实测积雪资料分析近年新疆积雪特征[J].高原气象,2008,27(3):551-557.
[1] 蒋玲梅,崔慧珍,王功雪,杨建卫,王健,潘方博,苏旭,方西瑶. 积雪、土壤冻融与土壤水分遥感监测研究进展[J]. 遥感技术与应用, 2020, 35(6): 1237-1262.
[2] 王世金,周蓝月,窦文康,颉佳,马兴刚. 基于GIS的不同时期冰川旅游服务潜力评价—以新疆维吾尔自治区为例[J]. 遥感技术与应用, 2020, 35(6): 1283-1291.
[3] 马腾耀,肖鹏峰,张学良,马威,郭金金. 基于特征优选的GF-3全极化数据积雪识别[J]. 遥感技术与应用, 2020, 35(6): 1292-1302.
[4] 雷华锦,李弘毅,王建,郝晓华,赵宏宇,张娟. 基于环境信息和回归模型的青藏高原MODIS积雪面积比例产品制备[J]. 遥感技术与应用, 2020, 35(6): 1303-1311.
[5] 乔海伟,张彦丽. 融合FY-3C号和FY-4A号卫星数据的积雪面积变化研究—以祁连山区为例[J]. 遥感技术与应用, 2020, 35(6): 1320-1328.
[6] 李毅,任超,张志刚,梁月吉,潘亚龙. 基于多元线性回归的GPS-IR积雪深度反演研究[J]. 遥感技术与应用, 2020, 35(6): 1312-1319.
[7] 徐瑾昊,冯敏,王建邦,冉有华,祁元,杨联安,李新. 基于高分遥感数据和深度学习的石冰川自动提取研究[J]. 遥感技术与应用, 2020, 35(6): 1329-1336.
[8] 杨瑞,祁元,苏阳. 深度学习U-Net方法及其在高分辨卫星影像分类中的应用[J]. 遥感技术与应用, 2020, 35(4): 767-774.
[9] 武黎黎,陈月庆,朱明,李晓峰,赵凯. 一种基于雪粒径演化过程的积雪亮温模拟新方法[J]. 遥感技术与应用, 2020, 35(3): 606-614.
[10] 张峰极,吴艳兰,姚雪东,梁泽毓. 基于改进DenseNet网络的多源遥感影像露天开采区智能提取方法[J]. 遥感技术与应用, 2020, 35(3): 673-684.
[11] 张正,肖鹏峰,张学良,冯学智,杨永可,胡瑞,盛光伟,刘豪. 青藏高原融雪期积雪反照率特性分析[J]. 遥感技术与应用, 2019, 34(6): 1146-1154.
[12] 陈秀雪,李晓峰,王广蕊,赵凯,郑兴明,姜涛. 基于积雪调查数据的东北地区被动微波积雪遥感产品精度验证与分析[J]. 遥感技术与应用, 2019, 34(6): 1181-1189.
[13] 李越帅,郑宏伟,罗格平,杨辽,王伟胜,桂东伟. 集成U-Net方法的无人机影像胡杨树冠提取和计数[J]. 遥感技术与应用, 2019, 34(5): 939-949.
[14] 刘天福,陈学泓,董琪,曹鑫,陈晋. 深度学习在GlobeLand30-2010产品分类精度优化中应用研究[J]. 遥感技术与应用, 2019, 34(4): 685-693.
[15] 林志玮,丁启禄,黄嘉航,涂伟豪,胡典,刘金福. 基于DenseNet的无人机光学图像树种分类研究[J]. 遥感技术与应用, 2019, 34(4): 704-711.