深度学习在GlobeLand30-2010产品分类精度优化中应用研究

doi:10.11873/j.issn.1004-0323.2019.4.0685

遥感技术与应用

2019, Vol. 34

Issue (4): 685-693 DOI: 10.11873/j.issn.1004-0323.2019.4.0685

CNN 专栏

深度学习在GlobeLand30-2010产品分类精度优化中应用研究

刘天福¹(

),陈学泓^1,²(

),董琪¹,曹鑫^1,²,陈晋^1,²

1. 地表过程与资源生态国家重点实验室北京师范大学地理科学学部，北京 100875
2. 北京市陆表遥感数据产品工程技术研究中心北京师范大学地理科学学部，北京 100875

Application of Deep Learning in GlobeLand30-2010 Product Refinement

Tianfu Liu¹(

),Xuehong Chen^1,²(

),Qi Dong¹,Xin Cao^1,²,Jin Chen^1,²

1. State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China
2. Beijing Engineering Research Center for Globe Land Remote Sensing Products, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China

全文: PDF(8047 KB) HTML

摘要：

本文提出结合深度卷积神经网络与在线高分遥感影像的分类方法，用于GlobeLand30地表覆盖产品的质量优化。首先，通过对多源地表覆盖产品的一致性分析，构建深度学习训练所需的高分辨率遥感大样本（224万样本量）；其次，基于该大规模样本集训练适用于GlobeLand30优化的深度卷积神经网络模型（GoogleNet Inception V3）；最后，利用训练好的神经网络模型对在线高分影像进行分类，用以优化GlobeLand30产品的不可靠区域。经独立测试样本集验证，经过训练的神经网络分类总体精度为87.7%，Kappa系数为0.86，相比原始GlobeLand30的精度（总体精度75.1%、Kappa系数0.71）有了明显提升。在4个试验区的GlobeLand 30产品优化实验表明：该方法能够有效优化GlobeLand30产品的分类精度。

关键词： 深度学习; GlobeLand30; 产品优化; Google Earth

Abstract:

GlobeLand30, as one of the best Globe Land Cover (GLC) product at 30 m resolution, was developed by China based on the integration of pixel- and object-based methods with knowledge (POK-based approach), which combines multiple levels of classification techniques to achieve high-accuracy land cover mapping. In particular, a knowledge-based verification process was employed to refine and grantee the product quality of Globeland30 by manual interpretation of online high-resolution images. However, the manual intervention suffers from large labor consumptions and the subjectivity influence. Considering the great achievements of deep learning in image recognition and classification, classifying online high-resolution remote sensing images with Deep Convolutional Neural Network (DCNN) may improve the efficiency and performance of the refinement procedure for GlobeLand30. However, the training of DCNN relies on a large number of training samples; and the existing remote sensing sample sets cannot satisfy the training requirements in terms of sample size and category system. Therefore, a method for generating large sample set of high-resolution remote sensing imagery based on multi-source GLC was proposed; and a large sample set with 2.24 million samples is automatically generated by this method. The DCNN model (GoogleNet Inception V3) was trained from scratch with the proposed large sample set and then used to refine Globeland30 product. Verification with an independent test sample set shows that the proposed trained DCNN model can achieve higher classification accuracy (Overall accuracy: 87.7%, Kappa: 0.856) than that of original GlobeLand30 product (Overall Accuracy: 75.1%, Kappa: 0.71). Finally, four test areas were selected for evaluating the performance of proposed refinement procedure. The results show that the GoogleNet (InceptionV3) model trained by the proposed large sample set can effectively refine the product quality of GlobeLand30.

Key words: Deep Learning GlobeLand30 Product Refinement Google Earth

收稿日期: 2018-10-19 出版日期: 2019-10-16

ZTFLH:

TP79

基金资助: 国家自然科学基金项目(41871224)

通讯作者: 陈学泓 E-mail: liutianfu@mail.bnu.edu.cn;chenxuehong@bnu.edu.cn

作者简介: 刘天福（1992-），男，江西赣州人，硕士研究生，主要从事遥感大数据制图与分析方面的研究。E?mail:liutianfu@mail.bnu.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
	刘天福
	陈学泓
	董琪
	曹鑫
	陈晋

引用本文:

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.

Tianfu Liu,Xuehong Chen,Qi Dong,Xin Cao,Jin Chen. Application of Deep Learning in GlobeLand30-2010 Product Refinement. Remote Sensing Technology and Application, 2019, 34(4): 685-693.

链接本文:

http://www.rsta.ac.cn/CN/10.11873/j.issn.1004-0323.2019.4.0685 或 http://www.rsta.ac.cn/CN/Y2019/V34/I4/685

图1 GlobeLand30-2010产品分类精度优化流程图

表1 现有部分全球地表覆盖数据集

表2 统一分类体系规则

图2 GLC联合数据与GlobeLand30匹配过程

图3 分类可靠和不可靠区域

图4 本文获取的高分辨率遥感样本示例集

图5 训练过程训练精度、验证精度和总体损失函数

表3 GlobeLand30-2010分类结果混淆矩阵

表4 DCNN模型分类结果混淆矩阵

图6 4个试验区的空间分布

图7 不同试验区优化结果

1	TurnerB L, LambinE F, ReenbergA. The Emergence of Land Change Science for Global Environmental Change and Sustainability[J]. Proceedings of the National Academy of Sciences, 2007, 104(52): 20666-20671.
2	ChenJ, ChenJ, LiaoA, et al. Global Land Cover Mapping at 30 m Resolution: A POK-based Operational Approach[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2015, 103: 7-27.
3	MaX, TongX, LiuS, et al. Optimized Sample Selection in SVM Classification by Combining with DMSP-OLS, Landsat NDVI and GlobeLand 30 Products for Extracting Urban Built-up Areas[J]. Remote Sensing, 2017, 9(3): 236.doi: 10.3390/rs9030236. doi: 10.3390/rs9030236
4	ChenJun, ChenJin, LiaoAnping, et al. Concepts and Key Techniques for 30 m Global Land Cover Mapping[J]. Acta Geodaetica et Cartographica Sinica, 2014, 43(6):551-557.
4	陈军, 陈晋, 廖安平, 等. 全球30 m地表覆盖遥感制图的总体技术[J]. 测绘学报, 2014, 43(6): 551-557.
5	ZhangWeiwei, ChenJun, LiaoAnping, et al. Geospatial Knowledge-based Verification and Improvement of Globeland30[J]. Science China Earth Sciences, 2016, 46(9): 1149-1161.
5	张委伟, 陈军, 廖安平, 等. 顾及多元知识的 GlobeLand30检核优化模型[J]. 中国科学：地球科学, 2016, 46(9): 1149-1161.
6	YangY, XiaoP, FengX, et al. Accuracy Assessment of Seven Global Land Cover Datasets over China[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2017, 125: 156-173.
7	HuangX, LiQ, LiuH, et al. Assessing and Improving the Accuracy of GlobeLand 30 Data for Urban Area Delineation by Combining Multisource Remote Sensing Data[J]. IEEE Geoscience and Remote Sensing Letters, 2016, 13(12): 1860-1864.
8	FonteC, MinghiniM, PatriarcaJ, et al. Generating Up-to-Date and Detailed Land Use and Land Cover Maps Using Open Street Map and GlobeLand30[J]. ISPRS International Journal of Geo-Information, 2017, 6(4): 125-147.
9	ChenDi, WuWenbin, LuMiao, et al. Progresses in Land Cover Data Reconstruction Method based on Multi-source Data Fusion[J]. Chinese Journal of Agricultural Resources and Regional Planning, 2016, 37(9): 62-70.
9	陈迪, 吴文斌, 陆苗, 等. 基于多源数据融合的地表覆盖数据重建研究进展综述[J]. 中国农业资源与区划, 2016 (9): 62-70.
10	YuLe, WangJie, LiXuecao, et al. A Multi-resolution Global Land Cover Dataset through Multi-source Data Aggregation[J]. Science China Earth Sciences, 2014, 44(8): 1646-1660.
10	俞乐, 王杰, 李雪草, 等. 基于多源数据集成的多分辨率全球地表覆盖制图[J]. 中国科学: 地球科学, 2014, 44(8): 1646-1660.
11	SeeL, SchepaschenkoD, LesivM, et al. Building a Hybrid Land Cover Map with Crowdsourcing and Geographically Weighted Regression[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2015, 103: 48-56.
12	FritzS, McCallumI, SchillC, et al. Geo-Wiki. Org: The Use of Crowdsourcing to Improve Global Land Cover[J]. Remote Sensing, 2009, 1(3): 345-354.
13	LeCunY, BengioY, HintonG. Deep Learning[J]. Nature, 2015, 521(7553): 436.
14	HeHaiqing, PangYan, ChenXiaoyong. Remote Sensing Image Scene Oriented Convolutional Neural Network Recursive Recognition Model[J]. Remote Sensing Technology and Application,2017,32(6):1078-1082.
14	何海清, 庞燕, 陈晓勇. 面向遥感影像场景的深度卷积神经网络递归识别模型[J]. 遥感技术与应用, 2017, 32(6): 1078-1082.
15	TianDeyu, ZhangYaonan, ZhaoGuohui, et al. Convolutional Neural Network for Remote Sensing Plant Cover Extracting[J]. Remote Sensing Technology and Application,2018,33(1):151-157.
15	田德宇, 张耀南, 赵国辉, 等. 基于卷积神经网络的遥感沙漠绿地提取方法[J]. 遥感技术与应用, 2018, 33(1): 151-157.
16	SchmidhuberJ. Deep Learning in Neural Networks: An Overview[J]. Neural Networks, 2015, 61: 85-117.
17	HuJ, JiangT, TongX, et al. A Benchmark for Scene Classification of High Spatial Resolution Remote Sensing Imagery[C]//2015 IEEE International Geoscience and Remote Sensing Symposium (IGARSS). IEEE, 2015: 5003-5006.
18	ChengG, HanJ, LuX. Remote Sensing Image Scene Classification: Benchmark and State of the Art[J]. Proceedings of the IEEE, 2017, 105(10): 1865-1883.
19	ZhouW, NewsamS, LiC, et al. PatternNet: A Benchmark Dataset for Performance Evaluation of Remote Sensing Image Retrieval[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2018, 145: 197-209.
20	NogueiraK, PenattiO A B, dos SantosJ A. Towards Better Exploiting Convolutional Neural Networks for Remote Sensing Scene Classification[J]. Pattern Recognition, 2017, 61: 539-556.
21	HuF, XiaG S, HuJ, et al. Transferring Deep Convolutional Neural Networks for the Scene Classification of High-resolution Remote Sensing Imagery[J]. Remote Sensing, 2015, 7(11): 14680-14707.
22	TsendbazarN E, de BruinS, HeroldM. Integrating Global Land Cover Datasets for Deriving User-specific Maps[J]. International Journal of Digital Earth, 2017, 10(3): 219-237.
23	CanzianiA, PaszkeA, CulurcielloE. An Analysis of Deep Neural Network Models for Practical Applications[J]. arXiv Preprint arXiv:1605.07678, 2016.
24	ZhangH, ZhangL, LiP. Fabric Pattern Classification based on Depth Convolution Neural Network[J]. Basic Sciences Journal of Textile Universities, 2017 (2): 361-265,267.
25	SzegedyC, VanhouckeV, IoffeS, et al. Rethinking the Inception Architecture for Computer Vision[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2016: 2818-2826.
26	ShenZ, LiuZ, LiJ, et al. Dsod: Learning Deeply Supervised Object Detectors from Scratch[C]//Proceedings of the IEEE International Conference on Computer Vision. 2017: 1919-1927.
27	SongX, HerranzL, JiangS. Depth CNNs for RGB-D Scene Recognition: Learning from Scratch Better than Transferring from RGB-CNNs[C]//Thirty-First AAAI Conference on Artificial Intelligence. 2017.
28	YiD, LeiZ, LiaoS, et al. Learning Face Representation from Scratch[J]. arXiv Preprint arXiv:1411.7923, 2014.

[1]	徐梦竹, 徐佳, 邓鸿儒, 袁春琦. 基于全极化SAR影像的海岛地物分类[J]. 遥感技术与应用, 2019, 34(3): 647-654.
[2]	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.
[3]	郝斌飞，韩旭军，马明国，刘一韬，李世卫. Google Earth Engine在地球科学与环境科学中的应用研究进展[J]. 遥感技术与应用, 2018, 33(4): 600-611.
[4]	张滔，唐宏. 基于Google Earth Engine的京津冀2001~2015年植被覆盖变化与城镇扩张研究[J]. 遥感技术与应用, 2018, 33(4): 593-599.
[5]	何海清，庞燕，陈晓勇. 面向遥感影像场景的深度卷积神经网络递归识别模型[J]. 遥感技术与应用, 2017, 32(6): 1078-1082.
[6]	王静璞,刘连友,沈玲玲. 基于Google Earth的毛乌素沙地新月形沙丘移动规律研究[J]. 遥感技术与应用, 2013, 28(6): 1094-1100.
[7]	朱睿,刘勇,尹振良,蒋红,许宝荣. Mashup在三维气象灾情预警信息系统中的应用[J]. 遥感技术与应用, 2012, 27(2): 322-326.
[8]	焦俊超,马安青,娄安刚,陈爽,杨少磊. GIS和Google Earth开发在溢油预测中的整合应用[J]. 遥感技术与应用, 2011, 26(2): 215-219.

Viewed

Full text

Abstract

Cited

Shared

Discussed