Please wait a minute...
img

官方微信

遥感技术与应用  2021, Vol. 36 Issue (2): 304-313    DOI: 10.11873/j.issn.1004-0323.2021.2.0304
CNN 专栏     
基于多时序特征和卷积神经网络的农作物分类
屈炀1,2(),袁占良1,赵文智2(),陈学泓2,陈家阁3
1.河南理工大学测绘与国土信息工程学院,河南 焦作 454003
2.北京师范大学地理科学学部,北京 100875
3.国家基础地理信息中心,北京 100830
Crop Classification based on Multi-temporal Features and Convolutional Neural Network
Yang Qu1,2(),Zhanliang Yuan1,Wenzhi Zhao2(),Xuehong Chen2,Jiage Chen3
1.School of Surveying and Land Information Engineering,Henan Polytechnic University,Jiaozuo 454003,China
2.Faculty of Geographical Science,Beijing Normal University,Beijing 100875,China
3.National Geomatics Center of China,Beijing 100830,China
 全文: PDF(5993 KB)   HTML
摘要:

近年来,以卷积神经网络为主的深度学习模型在各种遥感应用中都显示出巨大的潜力。以加州帝国郡为研究区,以Landsat 8 OLI年内时序遥感影像计算时序植被指数NDVI、EVI、RVI以及TVI,组合后输入到构建的一维卷积神经网络 模型,以实现作物的高精度精细分类。为了验证卷积模型的优越性,另搭建了基于递归神经网络及其变体的深度学习模型。结果表明:①引入其他时序特征后,能够有效地提高卷积神经网络的分类精度。NDVI+EVI+TVI+RVI组合特征总体精度和Kappa系数最高,分别是89.667 4%和0.856 0,对比NDVI时序特征总体精度和Kappa系数提高了近4%和0.6。②在与其他深度学习模型的对比中,一维卷积神经网络分类精度最高,能够从时序数据中较为准确捕捉作物时序特征信息,尽管递归神经网络被广泛应用于序列数据的研究,但分类结果要略差于卷积神经网络。实验表明在NDVI的基础上引入其他植被指数辅助,能够有效地提高分类精度。基于一维卷积神经网络的深度学习框架为长时间序列分类任务提供了一种有效且高效的方法。

关键词: 农作物分类一维卷积神经网络时间序列植被指数Landsat 8 OLI    
Abstract:

Recently, Convolutional Neural Network (CNN) shows great potential in various remote sensing applications Taking Imperial County of California as the study area, and calculating vegetation index NDVI, EVI, RVI and TVI form landsat-8 OLI time series remote sensing images. Then, input it into the constructed CNN model to achieve crop classification. In order to verify the superiority of the convolution model, a deep learning model based on recurrent neural networks and its variants was built. The results show that: ①Adding other time series features can effectively improve the classification accuracy of CNN. The overall accuracy and Kappa coefficient of NDVI+EVI+TVI+RVI combination features are best, respectively 89.6674% and 0.8560, which is nearly 4% and 0.6 higher than the single time series features. ②Convolutional neural networks have the highest classification accuracy in comparison with other deep learning models. It can capture crop timing feature information more accurately from time series data. Although RNN is widely used for sequential data representation, but the classification results are slightly worse than the convolutional neural network. Experiments show that the introduction of other vegetation index assistance on the basis of NDVI can effectively improve the classification accuracy. A deep learning framework based on 1D convolutional neural networks provides an effective and efficient method for Multi-Temporal classification tasks.

Key words: Crop classification    CNN    Time series    Vegetation index    Landsat 8 OLI
收稿日期: 2019-10-25 出版日期: 2021-05-24
ZTFLH:  TP79  
基金资助: 国家自然科学基金项目(41572341)
通讯作者: 赵文智     E-mail: quyang95@126.com;wenzhi.zhao@bnu.edu.cn
作者简介: 屈炀(1995-),男,河南许昌人,硕士研究生,主要从事遥感信息识别与提取研究。 E?mail:quyang95@126.com
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  
屈炀
袁占良
赵文智
陈学泓
陈家阁

引用本文:

屈炀,袁占良,赵文智,陈学泓,陈家阁. 基于多时序特征和卷积神经网络的农作物分类[J]. 遥感技术与应用, 2021, 36(2): 304-313.

Yang Qu,Zhanliang Yuan,Wenzhi Zhao,Xuehong Chen,Jiage Chen. Crop Classification based on Multi-temporal Features and Convolutional Neural Network. Remote Sensing Technology and Application, 2021, 36(2): 304-313.

链接本文:

http://www.rsta.ac.cn/CN/10.11873/j.issn.1004-0323.2021.2.0304        http://www.rsta.ac.cn/CN/Y2021/V36/I2/304

图1  研究区(2018年)
日期含云量/%日期含云量/%
2018.01.050.132018.06.300.00
2018.02.060.102018.08.017.13
2018.02.2211.122018.08.173.02
2018.03.266.882018.09.020.41
2018.04.115.642018.10.048.99
2018.04.270.002018.10.204.50
2018.05.130.642018.11.050.10
2018.05.295.212018.11.2114.18
表1  Landsat数据
图2  技术流程图
植被指数计算公式
NDVIρnir-ρredρnir-ρred
EVIG×ρnir-ρredρnir+C1ρred-C2ρblue+L
TVI0.5[120ρnir-ρg-200(ρred-ρgreen)]
RVI?ρnirρred
表2  采用的植被指数及其计算公式
图3  一维卷积模型结构
图4  不同特征组合分类结果
类型洋葱苜蓿冬小麦甜菜其他
NDVI特征卷积神经网络洋葱13 4611094451 0353 953
苜蓿030 39906111 061
冬小麦1524889978771 135
甜菜1322 195324 4415 444
其他1833 4492302 16959 855
制图精度/%97.6183.5192.9983.8983.77
用户精度/%70.8394.7979.8275.8790.85
总体精度:85.480 7 Kappa系数:0.798 6
NDVI+EVI特征卷积神经网络洋葱13 3382311929934 073
苜蓿031 11227847909
冬小麦663309 186776964
甜菜2239124125 1844 098
其他1833 8152291 34061 404
制图精度/%96.5885.4794.9586.4285.94
用户精度/%70.8594.5881.1382.6891.69
总体精度/%:87.381 7 Kappa系数:0.824 5

NDVI+TVI特征卷积神经

网络

洋葱13 5976393514362 895
苜蓿031 18105014 299
冬小麦114139 2005411 328
甜菜336381326 4454 100
其他1693 5341111 19058826
制图精度/%98.4685.6695.0990.7582.33
用户精度/%75.7986.6680.0584.6892.16
总体精度/%:86.774 1 Kappa系数:0.817 0

NDVI+EVI+TVI特征卷积

神经网络

洋葱13 6536023414442 585
苜蓿031 2420682484
冬小麦12599 2385122 156
甜菜211 214726 1364 406
其他1353058891 36661 817
制图精度/%98.8685.9095.4889.6986.52
用户精度/%77.4696.4075.9382.2393.01
总体精度/%:88.557 6 Kappa系数:0.841 0

NDVI+EVI+TVI+RVI

特征卷神经网络

洋葱13 6622413266563 344
苜蓿031 4410111547
冬小麦04559 302409989
甜菜3600126 8043 885
其他1453 66346116062 683
制图精度/%98.9386.3896.1491.9887.73
用户精度/%74.9597.9583.3985.6592.59
总体精度/%:89.667 4 Kappa系数:0.856 0
表3  特征组合分类精度对比
类型洋葱苜蓿冬小麦甜菜其他
递归神经网络(RNN)洋葱12 9759313756083 860
苜蓿528 98206131 603
冬小麦3411498 127876924
甜菜1591 99819925 0442 899
其他3304 3409741 99962 162
制图精度/%93.9579.6284.0085.9487.00
用户精度/%69.2092.8878.0282.6689.05
总体精度/%:85.553 3 Kappa系数:0.797 7
长短期记忆网络(LSTM)洋葱13 2069271116963 570
苜蓿4430 41020256794
冬小麦1153089 2668451 502
甜菜2679332824 6692 440
其他1783 8222502 67463 142
制图精度/%95.6383.5495.7784.6688.37
用户精度/%71.3596.4776.9987.0690.12
总体精度/%:87.673 9 Kappa系数:0.827 5
GRU洋葱13 1642482636394 747
苜蓿230 33705521 008
冬小麦1432368 8337521 038
甜菜3282567925 0432 102
其他1735 3235002 15462 553
制图精度/%95.3283.3491.3085.9487.55
用户精度/%69.0695.1080.2990.0688.47
总体精度/%:87.198 5 Kappa系数:0.820 4
表4  不同分类方法的分类精度
图5  不同分类方法的分类结果
1 Liu Jiyuan ,Shao Quanqin ,Yan Xiaodong, et al. Geobiophysical Effects of Land Use Change on Climate Change[J]. Chinese Journal of Nature,2014,36(5):356-363.
1 刘纪远,邵全琴,延晓冬,等.土地利用变化影响气候变化的生物地球物理机制[J].自然杂志,2014,36(5):356-363.
2 Shi Zhou, Liang Zongzheng ,Yang Yuanyuan, et al. Status and Prospect of Agricultural Remote Sensing[J]. Transactions of the Chinese Society for Agricultural Machinery , 2015,46(2):247-260.
2 史舟,梁宗正,杨媛媛,等.农业遥感研究现状与展望[J].农业机械学报,2015,46(2):247-260.
3 Tian Guoliang. Present Situation, Problems and Suggestions of Remote Sensing Application in China[J]. Remote Sensing Information , 2003(2):2-5.田国良.我国遥感应用现状、问题与建议[J].遥感信息,2003(2):2-5.
4 Wang Wenjing, Zhang Xia, Zhao Yindi,et al. Cotton Extraction Method of Integrated Multi-features based on Multi-temporal Landsat 8 Images[J]. Journal of Remote Sensing, 2017,21(1): 115–124
4 王文静,张霞,赵银娣,等.综合多特征的Landsat 8时序遥感图像棉花分类方法[J].遥感学报,2017,21(1):115-124.
5 Zhang Rongqun, Wang Sheng'an, Gao Wanlin ,et al. Remote Sensing Classification Method of County-level Agricultural Crops Using Time-series NDVI[J]. Transactions of the Chinese Society for Agricultural Machinery , 2015,46(S1):246-252.
5 张荣群,王盛安,高万林,等.基于时序植被指数的县域作物遥感分类方法研究[J].农业机械学报,2015,46():246-252.
6 Gu Xianghui,Zhang Ying,Sang Huiyong,et al.Research on Crop Classification Method based on Sentinel-2 Time Series Combined Vegetation Index[J].Remote Sensing Technology and Application,2020,35(3):702- 711.
6 谷祥辉,张英,桑会勇,等.基于哨兵 2 时间序列组合植被指数的作物分类研究[J]. 遥感技术与应用,2020,35(3):702-711.
7 Xue Huazhu,Wang Changjing,Zhou Hongmin,et al.BP Neural Network based on Simulated Annealing Algorithm for High Resolution LAI Retrieval[J].Remote Sensing Technology and Application,2020,35(5):1057- 1069.
7 薛华柱,王昶景,周红敏,等. 基于模拟退火算法的 BP 神经网络模型估算高分辨率叶面积指数[J]. 遥感技术 与应用,2020,35(5):1057-1069.
8 Wang Yingjie, Liu Liangyun, Wang Zhihui. Land Covermapping based on Landsat Time-series Stacks in Sanjiang Plain [J], Remote Sensing Technology and Application , 2015,30(5):959-968.
8 王颖洁,刘良云,王志慧.基于时序Landsat数据的三江平原植被地表类型变化遥感探测研究[J].遥感技术与应用,2015,30(5):959-968.
9 Ma Li, Xu Xingang, Jia Jianhua, et al. Crop Classification Method Using Multi-temporal TM Images[J]. Transactions of the Chinese Society of Agricultural Engineering, 2008, 24(S2):191-195.
9 马丽,徐新刚,贾建华,等.利用多时相TM影像进行作物分类方法[J].农业工程学报,2008,24():191-195.
10 Liang S , Fang H. Leaf Area Index Models☆[J]. Encyclopedia of Ecology, 2008, 77(5):2139-2148.
11 Guo Yushan, Liu Qingsheng, Liu Gaohuan,et al. Extraction of Main Crops in Yellow River Delta based on MODIS NDVI Time Series[J] Journal of Natural Resources, 2017,32(10):1808-1818.
11 郭昱杉,刘庆生,刘高焕,等.基于MODIS时序NDVI主要农作物种植信息提取研究[J].自然资源学报,2017,32(10):1808-1818.
12 Li Xiaohui, Wang Hong, Li Xiaobing,et al. Study on Crops Remote Sensing Classification based on Multi-temporal Landsat 8 OLI Images[J], Remote Sensing Technology and Application , 2019,34(2):389-397.
12 李晓慧,王宏,李晓兵,等.基于多时相Landsat 8 OLI影像的农作物遥感分类研究[J].遥感技术与应用,2019,34(2):389-397.
13 Wu Jing, Yu’na Lü , Li Chunbin, et al. Fine Classification of County Crops based on Multi-temporal Images of Sentinel-2A[J], Transactions of the Chinese Society for Agricultural Machinery, 2019,50(9):194-200.
13 吴静,吕玉娜,李纯斌,李全红.基于多时相Sentinel-2A的县域农作物分类[J].农业机械学报,2019,50(9):194-200.
14 Su Tengfei, Liu Quanming, Su Xiuchuan, Crop Remote Sensing Classification based on Multi Vegetation iIdex Time Series and Machine Learning[J], Jiangsu Agricultural Sciences, 2017,45(16):219-224.
14 苏腾飞,刘全明,苏秀川.基于多种植被指数时间序列与机器学习的作物遥感分类研究[J].江苏农业科学,2017,45(16):219-224.
15 Zhao W , Du S. Spectral-spatial Feature Extraction for Hyperspectral Image Classification: A Dimension Reduction and Deep Learning Approach[J]. IEEE Transactions on Geoscience and Remote Sensing, 2016, 54(8):4544-4554.
16 Li Yandong, Hao Zongbo, Lei Hang. Survey of Convolutional Neural Network[J]. Journal of Computer Applications, 2016,36(9):2508-2515,2565.
16 李彦冬,郝宗波,雷航.卷积神经网络研究综述[J].计算机应用,2016,36(9):2508-2515,2565.
17 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.
17 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.
18 Zhou Zhuang, Li Shengyang, Zhang Kang, et al.Crop Mapping Using Remotely Sensed Spectral and Context Features based on CNN[J] Remote Sensing Technology and Application, 2019,34(4):694-703.
18 周壮,李盛阳,张康,等.基于CNN和农作物光谱纹理特征进行作物分布制图[J].遥感技术与应用,2019,34(4):694-703.
19 Ienco D , Gaetano R , Dupaquier C , et al. Land Cover Classification via Multitemporal Spatial Data by Deep Recurrent Neural Networks[J]. IEEE Geoscience and Remote Sensing Letters, 2017:14(10):1685-1689.
20 Ndikumana E, Ho Tong Minh D, Baghdadi N, et al. Deep Recurrent Neural Network for Agricultural Classification Using Multitemporal SAR Sentinel-1 for Camargue, France[J]. Remote Sensing, 2018, 10(8): 1217. .
21 Interdonato R, Ienco D, Gaetano R, et al. DuPLO: A DUal View Point Deep Learning Architecture for Time Series ClassificatiOn[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2019, 149:91-104.
22 Xie Yi, Zhang Yongqing, Xun Lan, et al. Crop Classification based on Multi-source Remote Sensing Data Fusion and LSTM Algorithm[J]. Transactions of the Chinese Society of Agricultural Engineering, 2019, 35(15): 129-137.
22 解毅,张永清,荀兰,等. 基于多源遥感数据融合和 LSTM 算法的作物分类研究[J]. 农业工程学报,2019,35(15):129-137.
23 Belgiu M, Csillik O. Sentinel-2 Cropland Mapping Using Pixel-based and Object-based Time-weighted Dynamic Time Warping Analysis[J]. Remote Sensing of Environment,2018,204:509-520.
24 Chu QingWei, Zhang HongQun, Wu Yewei, et al. Application Research of Landsat-8[J]. Remote Sensing Information, 2013,28(4):110-114.
24 初庆伟,张洪群,吴业炜,等. Landsat 8卫星数据应用探讨[J].遥感信息,2013,28(4):110-114.
25 Boryan C, Yang Z, Mueller R, et al. Monitoring US Agriculture: the US Department of Agriculture[J]. National Agricultural Statistics Service, Cropland Data Layer Program. Geocarto Int,2011, 26: 341-358.
26 Hao Jianting, Yang Wunian, Li Yuxia, et al. Atmospheric Correction of Multi-spectral Imagery ASTER[J]. Remote Sensing Information, 2008(1):78-81.郝建亭,杨武年,李玉霞,等.基于FLAASH的多光谱影像大气校正应用研究[J].遥感信息,2008(1):78-81.
27 Yan L, Roy D P. Automated Crop Field Extraction from Multi-temporal Web Enabled Landsat Data[J]. Remote Sensing of Environment,2014,144:42-64.
28 Rouse J, Haas R, Schell J, et al. Monitoring Vegetation Systems in the Great Plains with ERTS[R]. Third ERTS Symposium, NASA (1973): 309-317.
29 Wang Zhengxing, Liu Chuang , Chen Wenbo. Preliminary Comparison of MODIS-NDVI and MODIS-EVI in Eastern Asia [J]. Geomatics and Information Science of Wuhan University, 2006(5):407-410,427.王正兴,刘闯,陈文波,等.MODIS增强型植被指数EVI与NDVI初步比较[J].武汉大学学报(信息科学版),2006(5):407-410,427.
30 Chen Xueyang, Meng Jihua, Du Xin, et al. The Monitoring of the Winter Wheat Leaf Area Index based on HJ-1 CCD Dada[J]. Remote Sensing for Land & Resources, 2010(2):55-58,62.陈雪洋,蒙继华,杜鑫,等.基于环境星CCD数据的冬小麦叶面积指数遥感监测模型研究[J].国土资源遥感,2010(2):55-58,62.
31 Broge N H,Leblanc E.Comparing Prediction Power and Stabi-lity of Broadband and Hyperspectral Vegetation Indices for Estimation of Green Leaf Area Index and Canopy Chlorophyll Den-sity[J].Remote Sensing of Environment,2001,76(2):156-172.
32 Bian Jinhu,Li Ai'nong,Song Mengqiang,et al. Reconstruction of NDVI Time-series Datasets of MODIS based on Savitzky-Golay Filter[J]. Journal of Remote Sensing, 2010,14(4):725-741.
32 边金虎,李爱农,宋孟强,等.MODIS植被指数时间序列Savitzky-Golay滤波算法重构[J].遥感学报,2010,14(4):725-741.
[1] 郭利彪,刘桂香,运向军,张勇,孙世贤. 基于数据机理的植被叶面积指数遥感反演研究[J]. 遥感技术与应用, 2020, 35(5): 1047-1056.
[2] 唐希颖,崔耀平,李楠,付一鸣,刘小燕,闰亚迪. 2000~2015年北京市土地利用强度及其辐射反馈评估[J]. 遥感技术与应用, 2020, 35(3): 587-595.
[3] 王红岩,汪晓帆,高亮,李强子,赵龙才,杜鑫,张源. 基于季相变化特征的撂荒地遥感提取方法研究[J]. 遥感技术与应用, 2020, 35(3): 596-605.
[4] 董超,赵庚星. 时序数据集构建质量对土地覆盖分类精度的影响研究[J]. 遥感技术与应用, 2020, 35(3): 558-566.
[5] 岳昂,曾庆伟,王怀警. 于桥水库蓝藻水华遥感长时序监测研究[J]. 遥感技术与应用, 2020, 35(3): 694-701.
[6] 谷祥辉,张英,桑会勇,翟亮,李少军. 基于哨兵2时间序列组合植被指数的作物分类研究[J]. 遥感技术与应用, 2020, 35(3): 702-711.
[7] 柴旭荣,李明,周义,王金风,田庆春. 影像的土地覆被快速分类[J]. 遥感技术与应用, 2020, 35(2): 315-325.
[8] 张新平,乔治,李皓,闫杰,张芳芳,赵栋锋,王得祥,康海斌,杨航,冯扬. 基于Landsat影像和不规则梯形方法遥感反演延安城市森林表层土壤水分[J]. 遥感技术与应用, 2020, 35(1): 120-131.
[9] 王一帆,徐涵秋. 利用MODIS EVI时间序列数据分析福建省植被变化(2000~2017年)[J]. 遥感技术与应用, 2020, 35(1): 245-254.
[10] 刘新杰,魏云霞,焦全军,孙奇,刘良云. 基于时序定量遥感的冬小麦长势监测与估产研究[J]. 遥感技术与应用, 2019, 34(4): 756-765.
[11] 张海东,田婷,张青,陆洲,石春林,谭昌伟,罗明,钱春花. 基于GF-1影像的耕地地块破碎区水稻遥感提取[J]. 遥感技术与应用, 2019, 34(4): 785-792.
[12] 黄静,王芳,张运. 基于PSI技术的芜湖市地面沉降时空特征分析[J]. 遥感技术与应用, 2019, 34(4): 829-838.
[13] 云增鑫, 郑光, 马利霞, 王晓菲, 卢晓曼, 路璐. 联合主被动遥感数据定量评价林下植被对叶面积指数估算的影响[J]. 遥感技术与应用, 2019, 34(3): 583-594.
[14] 许文鑫, 周玉科, 梁娟珠. 基于变化点的青藏高原植被时空动态变化研究 [J]. 遥感技术与应用, 2019, 34(3): 667-676.
[15] 李晓慧, 王宏, 李晓兵, 迟登凯, 汤曾伟, 韩重远. 基于多时相Landsat 8 OLI影像的农作物遥感分类研究[J]. 遥感技术与应用, 2019, 34(2): 389-397.