Please wait a minute...
img

官方微信

遥感技术与应用  2019, Vol. 34 Issue (4): 748-755    DOI: 10.11873/j.issn.1004-0323.2019.4.0748
CNN 专栏     
深度学习对不同分辨率影像冬小麦识别的适用性研究
崔刚(),吴金胜(),于镇,周玲
国家统计局山东调查总队, 山东 济南 250001
Comparison Analysis on Wheat Mapping Using Deep Learning Algorithm from Different Satellite Data Source
Gang Cui(),Jinsheng Wu(),Zhen Yu,Ling Zhou
Survey Office of the National Bureau of Statistics in Shandong, Ji’nan, 250001, China
 全文: PDF(8929 KB)   HTML
摘要:

定量分析遥感影像尺度与分类精度之间的关系是进行土地覆盖分类的基础。深度学习具有从底层到高层特征非监督学习的能力,解决了传统分类模型中需要人工选择特征的问题。这种新型的分类方法的分类精度是否受到不同分辨率尺度影响,有待研究。利用深度卷积神经网络(Deep Convolutional Neural Network, DCNN)——金字塔场景分析网络(Pyramid Scene Parsing Network, PSPNet)进行4种分辨率(8、3.2、2和0.8 m)的米级、亚米级影像冬小麦分类。实验结果表明: PSPNet能够有效地进行大样本的学习训练,非监督提取出空间特征信息,实现“端—端”的冬小麦自动化分类。不同于传统分类器分类精度与分类尺度之间的关系,随着影像分辨率的逐步增高,地物表达特征越来越清晰,PSPNet识别的冬小麦精度会逐步增高,识别地块结果也越来越规整,不受分辨率尺度的影响。这对于选择甚高亚米级影像提高作物分类精度提供了实验基础。

关键词: 图像融合深度卷积神经网络ResNetPSPNet高分1/2号卫星    
Abstract:

Quantitative analysis on the relationships between the remote sensing scale and the land cover classification accuracy, which is the basis for making a decision on remote sensing resolution determination, is essential for mapping the concise land cover. Up to now, deep learning is an innovative algorithm to learn the hierarchical layer features without supervised control, which is different from the traditional classifiers that require man-made labels as input. Therefore, it is interesting to explore the inherent relationship between the classification accuracy and remote sensing image spatial scale from this algorithm. In this paper, we applied a Deep Convolutional Neural Network (DCNN) which is Pyramid Scene Parsing Network (PSPNet) on four scale remote sensing image (8 m, 3.2 m, 2 m, 0.8 m) to map the wheat distribution. The experiment results showed that the PSPNet is good at learning the spatial feature without manual operations, then the wheat extent could be extracted automatically. Different from the conventional algorithm of determining the optimized spatial resolution, the PSPNet could identify the wheat better accompanying with the spatial resolution increased and more concise wheat results could be obtained. This conclusions represent that deep convolution neural network can take full use of the spatial information of the high remote sensing image to ensure the performance of wheat extent, which brings us a new idea of improving the accuracy of crop mapping adequately if we can get the super-high resolution remote sensing image.

Key words: Image fusion    Deep Convolution Neural Network    ResNet    PSPNet    GF-1/2
收稿日期: 2018-05-10 出版日期: 2019-10-16
ZTFLH:  S127  
基金资助: 山东三农普无人机飞行测量服务项目
通讯作者: 吴金胜     E-mail: cuigang@stats;wjs@stats
作者简介: 崔 刚(1967-),男,山东茌平人,高级统计师,主要从事社会经济调查、农业调查。E?mail:cuigang@stats?sd.gov.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  
崔刚
吴金胜
于镇
周玲

引用本文:

崔刚,吴金胜,于镇,周玲. 深度学习对不同分辨率影像冬小麦识别的适用性研究[J]. 遥感技术与应用, 2019, 34(4): 748-755.

Gang Cui,Jinsheng Wu,Zhen Yu,Ling Zhou. Comparison Analysis on Wheat Mapping Using Deep Learning Algorithm from Different Satellite Data Source. Remote Sensing Technology and Application, 2019, 34(4): 748-755.

链接本文:

http://www.rsta.ac.cn/CN/10.11873/j.issn.1004-0323.2019.4.0748        http://www.rsta.ac.cn/CN/Y2019/V34/I4/748

图1  山东19个村的空间分布
图2  基于PSPNet的冬小麦识别流程
图3  影像与对应的融合影像(子区图)
图4  4种图像的影像光谱方差
生产精度/%用户精度/%F1-Score

总体精度

/%

PSPNetGF-1-MUL0.800.880.8483.57
GF-1-GS0.840.890.8685.26
GF-2-MUL0.900.850.8789.12
GF-2-GS0.920.860.8990.78
SVMGF-1-MUL0.770.830.8080.29
GF-1-GS0.810.860.8385.26
GF-2-MUL0.880.850.8686.67
GF-2-GS0.880.830.8584.35
表1  4种尺度冬小麦分类精度
图5  分辨率尺度冬小麦识别结果
1 MingDongping, WangQun, YangJianyu. Spatial Scale of Remote Sensing Image and Selection of Optimal Spatial Resolution[J]. Journal of Remote Sensing, 2008, 12(4): 529-537.
1 明冬萍,王群,杨建宇. 遥感影像空间尺度特性与最佳空间分辨率选择[J]. 遥感学报. 2008,12(4): 529-537.
2 MingDongping, QiuYufang, ZhouWen. Applying Spatial Statistics into Remote Sensing Pattern Recognition: with Case Study of Cropland Extraction based on GeOBIA[J]. Acta Geodaetica of Cartographica Sinica, 2016, 45(7): 825-833.
2 明冬萍, 邱玉芳, 周文.遥感模式分类中的空间统计学应用——以面向对象的遥感影像农田提取为例[J].测绘学报, 2016, 45(7): 825-833.
3 MarkhamB L, TownshendJ R G. Land Cover Classification Accuracy as a Function of Sensor Spatial Resolution[C]∥ Proceedings 15th Int. Symp.on Remote Sensing of Environment, Ann Arbor, MI, 1981.
4 MarceauD J, HayG J. Remote Sensing Contributions to the Scale Issue[J]. Canadian Journal of Remote Sensing, 1999, 25 (4): 357-366.
5 WoodcockC E, StrahlerA H. The Factor of Scale in Remote Sensing[J]. Remote Sensing of Environment, 1987, 21(3): 311-332.
6 MarkowitzH. Portfolio Selection-Efficient Diversification of Investments[M]. New York: Wiley, 1959.
7 ChenC L, WuG. Choice of Optimal Scale for Multi-source Remote Sensing Images[J]. Journal of Zhejiang A & F University, 2011,28(1): 164-172.
8 HanP, GongJ Y, LiZ L, et al. Selection of Optimal Scale in Remotely Sensed Image Classification[J]. Journal of Remote Sensing, 2010, 14(3): 507-518.
9 FengGuixiang, MingDongping. Fractal based Method on Selecting the Optimal Spatial Resolution for Remote Sensing Image[J]. Journal of Geo-Information Science, 2015, 17(4):478-495.
9 冯桂香, 明冬萍. 分形定量选择遥感影像最佳空间分辨率的方法与实验. 地理信息科学, 2015,17(4): 478-495.
10 YangYanjun, TianQingjiu, ZhanYulin, et al. Effects of Spatial Resolution and Texture Features on Multi-spectral Remote Sensing Classification[J]. Journal of Geo-information Science, 2018,20(1): 99-107.
10 杨闫君, 田庆久, 占玉林,等. 空间分辨率与纹理特征对多光谱遥感分类的影响. 地理信息科学, 2018,20(1): 99-107.
11 ReichsteinM, Camps-VallsG, StevensB, et al. Deep Learning and Process Understanding for Data-Driven Earth System Science[J]. Nature, 2019, 566(7743): 195-204.
12 MaggioriE, TarabalkaY, CharpiatG, et al. Convolutional Neural Networks for Large-scale Remote Sensing Image Classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2016, 55(2):645-657.
13 WeiY N, WangZ L, XuM. Road Structure Refined CNN for Road Extraction in Aerial Image[J]. IEEE Geoscience and Remote Sensing Letters, 2017, 14(5):709-713.
14 ZhangBing. Remotely Sensed Big Data Era and Intelligent Information Extraction[J]. Geomatics and Information Science of Wuhan University, 2018, 53(12): 1861-1871.
14 张兵. 遥感大数据时代与智能信息提取[J]. 武汉大学学报(信息科学版), 2018, 53(12): 1861-1871.
15 ZhangKang, BaoqinHei, LiShengyang, et al. Complex Scene Classification of Remote Sensing Images based on CNN [J]. Remote Sensing for Land Resources, 2018, 33(6): 1095-1102.
15 张康, 黑保琴, 李盛阳,等. 基于CNN模型的遥感图像复杂场景分类[J]. 国土资源遥感, 2018, 33(6): 49-55.
16 YokoyaN, IwasakiA. Object Detection based on Sparse Representation and Hough Voting for Sptical Remote Sensing Imagery[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2015, 8(5): 2053-2062.
17 XuKaijian, TianQingjiu, YangYanjun, et al. Response of Spatial Scale for Land Cover Classification of Remote Sensing[J]. Journal of Geo-Information Science, 2018,20(2):246-253.
17 徐凯健,田庆久,杨闫君,等.遥感土地覆被分类的空间尺度响应研究[J].地球信息科学学报,2018,20(2):246-253.
18 Zhao H S ShiJ P, QiX J, et al. Pyramid Scene Parsing Network[C]∥ Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017: 2881-2890.
19 KrizhevskyA, SutskeverI, HintonG E. ImageNet Classification with Deep Convolutional Neural Networks[C]∥ International Conference on Neural Information Processing Systems. Curran Associates Inc. 2012:1097-1105.
20 SimonyanK, ZissermanA. Very Deep Convolutional Networks for Large-Scale Image Recognition[J]. Computer Science, 2014:1409-1556.
21 SzegedyC, IoffeS, VanhouckeV, et al. Inception-v4, Inception-Resnet and the Impact of Residual Connections on Learning[C]∥ Thirty-First AAAI Conference on Artificial Intelligence, 2017, 4278-4284
22 HeK, ZhangX, RenS, et al. Deep Residual Learning for Image Recognition[C]∥ Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778.
23 YosinskiJ, CluneJ, BengioY, et al. How Transferable are Features in Deep Neural Networks?[C]∥ Advances in Neural Information Processing Systems, 2014: 3320-3328.
24 ZhouJingping, LiCunjun, ShiLeigang, et al. Crops Distribution Remote Sensing Extraction based on Decision Tree and Object-oriented Method[J]. Transactions of the Chinese Society for Agricultural Machinery, 2016, 47(9): 318-326.
24 周静平,李存军,史磊刚,等. 基于决策树和面向对象的作物分布信息遥感提取[J]. 农业机械学报,2016,47(9):318-326.
25 WengYongling, TianQingjiu. Analysis and Evaluation of Method on Remote Sensing Date Fusion[J]. Remote Sensing Information,2003(3):49-54.
25 翁永玲,田庆久.遥感数据融合方法分析与评价综述[J].遥感信息,2003(3):49-54.
26 LiShuming, FengQuanlong, LiangQichun, et al. Aircraft Auto-detection in Domestic High Resolution Remote Sensing Images Using Deep-learning[J]. Remote Sensing Technology and Application, 2018, 33(6): 1095-1102.
26 李淑敏, 冯权泷, 梁其椿, 张学庆.基于深度学习的国产高分遥感影像飞机目标自动检测[J].遥感技术与应用, 2018, 33(6): 1095-1102.
[1] 肖新耀,许宁,尤红建. 一种基于à trous小波和联合稀疏表示的遥感图像融合方法[J]. 遥感技术与应用, 2015, 30(5): 1021-1026.
[2] 别强,何磊,赵传燕. 基于影像融合和面向对象技术的植被信息提取研究[J]. 遥感技术与应用, 2014, 29(1): 164-171.
[3] 赖格英,曾祥贵,刘影,张玲玲,易发钊,潘瑞鑫,盛盈盈. 基于ETM和图像融合的优势植被冠层叶面积指数和消光系数的遥感反演[J]. 遥感技术与应用, 2013, 28(4): 697-706.
[4] 孙萍,邓磊,聂娟. 一种基于区域分割的多尺度遥感图像融合方法[J]. 遥感技术与应用, 2012, 27(6): 844-849.
[5] 王晓艳,刘勇,蒋志勇. 一种基于结构相似度的IHS变换融合算法[J]. 遥感技术与应用, 2011, 26(5): 670-676.
[6] 薛东剑,张东辉,何政伟,张雪峰. 多源遥感影像融合技术在地质灾害调查中的应用[J]. 遥感技术与应用, 2011, 26(5): 664-669.
[7] 陈超,秦其明,池长艳,蒋洪波,刘明超. 一种Curvelet变换和IHS变换相结合的遥感图像融合方法[J]. 遥感技术与应用, 2011, 26(4): 444-449.
[8] 唐王琴,梁栋,胡根生,马雪亮,杭丹萍. 基于支持向量机的遥感图像厚云去除算法[J]. 遥感技术与应用, 2011, 26(1): 111-116.
[9] 刘廷祥,黄丽梅,鲍文东. 基于CBERS-02B和SPOT-5全色波段的图像融合纹理信息评价研究[J]. 遥感技术与应用, 2009, 24(1): 103-108.
[10] 陈 雯,王远飞. 基于GA—BP算法的多分辨率遥感影像融合技术[J]. 遥感技术与应用, 2007, 22(4): 555-559.
[11] 黎新亮,赵书河,柯长青,管开宇. 遥感图像融合定量评价方法及实验研究[J]. 遥感技术与应用, 2007, 22(3): 460-465.
[12] 员永生, 杨为民. 应用ASTER 数据提取高程和树种信息的研究[J]. 遥感技术与应用, 2006, 21(4): 307-311.
[13] 张宁玉, 吴泉源. Brovey 融合与小波融合对QuickBird 图像的信息量影响[J]. 遥感技术与应用, 2006, 21(1): 67-70.
[14] 吴 樊,王 超,张卫国,张 红. 机载综合孔径辐射计图像和光学图像融合及评价[J]. 遥感技术与应用, 2005, 20(4): 425-429.
[15] 周前祥,敬忠良. 面向对象的遥感图像融合处理系统的设计与应[J]. 遥感技术与应用, 2004, 19(1): 15-19.