基于GlobeLand30的智能解译样本动态制备方法

doi:10.14088/j.cnki.issn0439-8114.2023.09.026

湖北农业科学 ›› 2023, Vol. 62 ›› Issue (9): 142-150.doi: 10.14088/j.cnki.issn0439-8114.2023.09.026

基于GlobeLand30的智能解译样本动态制备方法

陈静^1,2, 何翔宇³, 陈建胜¹, 陈静波¹, 邓毓弸^1,2, 张学华⁴

1.中国科学院空天信息创新研究院,北京 100101;
2.中国科学院大学电子电气与通信工程学院,北京 101400;
3.军委后勤保障部工程质量监督中心,北京 100142;
4.应急管理部国家减灾中心,北京 100124

收稿日期:2022-01-29 出版日期:2023-09-25 发布日期:2023-10-24
通讯作者: 陈建胜（1986-）,男,山东潍坊人,助理研究员,博士,主要从事遥感图像处理研究,（电话）13718739737（电子信箱）chenjs@aircas.ac.cn。
作者简介:陈静（1995-）,女,甘肃庆阳人,在读硕士研究生,研究方向为遥感图像处理,（电话）18511431833（电子信箱）2673663031@qq.com。
基金资助:
中国科学院空天信息创新研究院自主部署项目（E1Z211010F）

A dynamic preparation method for intelligent interpretation samples based on GlobeLand30

CHEN Jing^1,2, HE Xiang-yu³, CHEN Jian-sheng¹, CHEN Jing-bo¹, DENG Yu-peng^1,2, ZHANG Xue-hua⁴

1. Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100101, China;
2. School of Electronic, Electrical and Communication Engineering, University of Chinese Academy of Sciences, Beijing 101400, China;
3. Engineering Quality Supervision Center of Logistics Support Department of the Military Commission, Beijing 100142, China;
4. National Disaster Reduction Center of the Ministry of Emergency Management, Beijing 100124, China

Received:2022-01-29 Online:2023-09-25 Published:2023-10-24

摘要/Abstract

摘要： 从影像特征相似性出发研究基于Landsat-8影像相似性的智能解译样本动态制备方法,构造光谱相似度、纹理相似度、空间邻近度3种样本集选择方法。基于相同的U-Net+EfficientNet-B3语义分割网络,对比了根据3种测度下选择的样本影像所制备的样本集对总体分类精度的影响。结果表明,从历史成果数据中选择样本影像用于模型训练,是提高分类精度的有效方法;在3种样本影像选择策略中,空间邻近度可得到精度最高、方差最小的分类结果;历史成果数据中的错误标签会导致智能模型精度降低。

关键词: GlobeLand30, 智能解译样本, 动态制备方法, 语义分割, 数据集, Landsat-8, 影像相似性

Abstract: A dynamic preparation method for intelligent interpretation samples based on Landsat-8 image similarity was studied from the perspective of image feature similarity. Three sample set selection methods were constructed, including spectral similarity, texture similarity, and spatial proximity.Based on the same U-Net+EfficientNet-B3 semantic segmentation network, the impact of sample sets prepared from selected sample images under three measures on overall classification accuracy was compared. The results indicated that selecting sample images from historical achievement data for model training was an effective method to improve classification accuracy;among the three sample image selection strategies, spatial proximity could obtain the classification results with the highest accuracy and the lowest variance;incorrect labels in historical achievement data could lead to a decrease in the accuracy of intelligent models.

Key words: GlobeLand30, intelligent interpretation samples, dynamic preparation method, semantic segmentation, dataset, Landsat-8, image similarity

中图分类号:

TP751

陈静, 何翔宇, 陈建胜, 陈静波, 邓毓弸, 张学华. 基于GlobeLand30的智能解译样本动态制备方法[J]. 湖北农业科学, 2023, 62(9): 142-150.

CHEN Jing, HE Xiang-yu, CHEN Jian-sheng, CHEN Jing-bo, DENG Yu-peng, ZHANG Xue-hua. A dynamic preparation method for intelligent interpretation samples based on GlobeLand30[J]. HUBEI AGRICULTURAL SCIENCES, 2023, 62(9): 142-150.

参考文献

[1] ANGEL S, PARENT J, CIVCO D L, et al.The dimensions of global urban expansion:Estimates and projections for all countries, 2000—2050[J]. Progress in planning, 2011, 75(2): 53-107.
[2] ELLIS E C, FULLER D Q, KAPLAN J O, et al.Dating the Anthropocene: Towards an empirical global history of human transformation of the terrestrial biosphere dating the anthropocene[J]. Elementa: Science of the anthropocene, 2013, 1.
[3] 陈军, 廖安平, 陈晋, 等.全球30 m地表覆盖遥感数据产品—GlobeLand30[J].地理信息世界,2017,24(1):1-8.
[4] CHEN B, XU B, ZHU Z, et al.Stable classification with limited sample: Transferring a 30 m resolution sample set collected in 2015 to mapping 10 m resolution global land cover in 2017[J]. Science bulletin, 2019, 64: 370-373.
[5] ZHANG X, LIU L Y, WU C S, et al.Development of a global 30 m impervious surface map using multisource and multitemporal remote sensing datasets with the Google Earth Engine platform[J]. Earth system science data, 2020,12(3): 1625-1648.
[6] 陈军, 陈晋, 廖安平.全球地表覆盖遥感制图[M]. 北京: 科学出版社, 2016.
[7] 刘晓曈. 基于深度学习的分类预测方法研究及应用[D] .南京:东南大学,2017.
[8] HANSEN M C, REED B.A comparison of the IGBP DISCover and university of maryland 1 km global land cover products[J]. International journal of remote sensing, 2000, 21(6-7): 1365-1373.
[9] HEROLD M, MAYAUX P, WOODCOCK C E, et al.Some challenges in global land cover mapping: An assessment of agreement and accuracy in existing 1 km datasets[J]. Remote sensing of environment, 2008, 112(5): 2538-2556.
[10] YU L, WANG J, GONG P.Improving 30 m global land-cover map FROM-GLC with time series MODIS and auxiliary data sets: A segmentation-based approach[J]. International journal of remote sensing, 2013, 34(16): 5851-5867.
[11] 党宇,张继贤,邓喀中,等.基于深度学习AlexNet的遥感影像地表覆盖分类评价研究[J].地球信息科学学报,2017,19(11):1530-1537.
[12] 朱明. 卷积神经网络在高分辨率卫星影像地表覆盖分类中的应用研究[D] .北京:中国地质大学,2020.
[13] TONG X Y, XIA G S, HU F, et al.Exploiting deep features for remote sensing image retrieval: A systematic investigation[J]. Ieee transactions on big data, 2019, 6(3): 507-521.
[14] KANU S, KHOJA R, LAL S,et al.CloudX-net: A robust encoder-decoder architecture for cloud detection from satellite remote sensing images[J].Remote sensing applications society and environment, 2020, 20:100417.
[15] SILVA F C N, LIMITED O B.Skyscapes: The Role and Importance of the sky in Archaeology[J].European journal of archaeology, 2015, 19(2):1-5.
[16] LI Y, SHAO Z, HUANG X,et al.Meta-FSEO: A meta-learning fast adaptation with self-supervised embedding optimization for few-shot remote sensing scene classification[J].Remote sensing, 2021, 13(14):2776.
[17] 刘洲洲. 面向不平衡数据集的对比模式挖掘算法研究[D].长沙:湖南大学,2016.
[18] ALHICHRI H, ALSUWAYED A, BAZI Y,et al.Classification of remote sensing images using efficientnet-B3 CNN model with attention[J].IEEE Access, 2021, 99:101-112.
[19] EICKEN T V, BASU A, BUCH V,et al.U-Net: A user-level network interface for parallel and distributed computing (includes URL)[J].ACM sigops operating systems review, 1995, 29(5):40-53.
[20] 云影. 美国地球观测卫星Landsat-8[J].卫星应用,2013(2):76.
[21] RONALD L G.An efficient algorithm for determining the convex hull of a finite planar set[J]. Information processing letters, 1972,1(4):132-133.
[22] 董晓莉, 顾成奎, 王正欧.基于形态的时间序列相似性度量研究[J].电子与信息学报, 2007, 29(5):1228-1231.
[23] TOBLER W.A computer movie simulating urban growth in the detroit region[J]. Economic geography,1970,46(2):234-240.
[24] LI Y S, ZHANG Y J, HUANG X, et al.Large-scale remote sensing image retrieval by deep hashing neural networks[J]. IEEE transactions on geoscience and remote sensing, 2017, 56(2): 950-965.

基于GlobeLand30的智能解译样本动态制备方法

A dynamic preparation method for intelligent interpretation samples based on GlobeLand30

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 3

编辑推荐

Metrics

本文评价

[1]	陈敏, 刘加敏, 秦亮军, 陈奕云. 基于GlobeLand30的广州市生态系统服务价值与社会经济协同发展格局分析[J]. 湖北农业科学, 2023, 62(9): 14-19.
[2]	刘震, 纪明妹, 郭志顶, 黄素芳, 赵忠祥, 闫旭东, 滕霄, 石秘, 岳明强, 刘青松, 徐玉鹏. 河北省农业害虫图像识别系统建设[J]. 湖北农业科学, 2022, 61(7): 135-139.
[3]	刘凯露, 王远弘, 买买提艾力·买买提依明, 杨帆, 刘永强. 基于FTIR野外实测的沙漠下垫面地表发射率数据集[J]. 湖北农业科学, 2021, 60(24): 197-202.