地球信息科学学报 ›› 2020, Vol. 22 ›› Issue (4): 772-783.doi: 10.12082/dqxxkx.2020.200192

• • 上一篇    下一篇


董金玮1,*(), 吴文斌2, 黄健熙3, 尤南山1, 何盈利1, 闫慧敏1   

  1. 1. 中国科学院地理科学与资源研究所,北京 100101
    2. 中国农业科学院农业资源与农业区划研究所,北京 100081
    3. 中国农业大学土地科学与技术学院,北京 100083
  • 收稿日期:2020-04-07 修回日期:2020-04-20 出版日期:2020-04-25 发布日期:2020-06-10
  • 通讯作者: 董金玮 E-mail:dongjw@igsnrr.ac.cn
  • 作者简介:董金玮(1982— ),男,山东潍坊人,研究员,博士生导师,主要从事土地利用与植被变化遥感研究。
  • 基金资助:

State of the Art and Perspective of Agricultural Land Use Remote Sensing Information Extraction

DONG Jinwei1,*(), WU Wenbin2, HUANG Jianxi3, YOU Nanshan1, HE Yingli1, YAN Huimin1   

  1. 1. Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China
    2. Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing 100081, China
    3. College of Land Science and Technology, Chinese Agricultural University, Beijing 100083, China
  • Received:2020-04-07 Revised:2020-04-20 Online:2020-04-25 Published:2020-06-10
  • Contact: DONG Jinwei E-mail:dongjw@igsnrr.ac.cn
  • Supported by:
    The Strategic Priority Research Program of Chinese Academy of Sciences(XDA19040301);National Natural Science Foundation of China(41871349)


农业用地占到全球土地面积近一半,农业土地利用(包括耕地及作物分布、种植制度、土地管理等)变化直接影响到粮食安全、水安全、生态安全和气候变化。遥感已经成为土地利用信息获取的重要手段,近年来中分辨率遥感卫星如Landsat、Sentinel以及中国高分卫星等的免费开放为国内外农业土地利用信息提取提供了前所未有的机遇,取得了一系列重要研究进展。本文从耕地分布、作物类型识别、农业种植制度以及农业土地管理4个角度分析了土地利用信息提取的最新研究进展。结果发现:① 耕地分布产品已经由过去的粗分辨率提升到10~30 m,耕地现状数据较为丰富,但挖掘遥感数据实现耕地变化历史回溯的能力有待加强;② 作物分类方面多采用地面调查数据和卫星遥感(Landsat和Sentinel-2为主)相结合的方式进行,在北美和欧洲得到了业务化运行,但对作物种植面积早期监测的能力有待加强;③ 基于遥感的农业种植制度信息获取(如撂荒)研究多集中在东欧等地区,在中国由于经济和政策因素导致的撂荒、轮作、休耕等现象也十分普遍,但具有针对性的遥感监测研究目前还相对缺乏;④ 农业土地管理措施信息提取方面,区域灌溉面积产品取得了重要进展,但数据的可靠性和准确性仍有待提高。在此基础上,我们结合遥感大数据、深度学习算法、云计算平台的发展对未来农业土地利用信息提取研究进行了展望:① 融合多源数据形成更高维度空间、光谱和时间信息的遥感大数据,提升特征提取和数据挖掘能力;② 机器学习和深度学习算法等智能化方法与基于地理学和物候信息的专家知识方法的耦合;③ 遥感云计算和大数据挖掘等前沿遥感和计算技术的应用。

关键词: 农业土地利用, 遥感, 耕地分布, 作物分类, 农业种植制度, 农业土地管理措施


Agricultural lands account for nearly half of the global land area, and changes in agricultural land use directly affect food security, water security, ecological security, and climate change. Remote sensing is the main means for acquiring agricultural land use information. In recent years, the free opening of medium-resolution remote sensing data such as Landsat, Sentinel, and China's GaoFen satellites has opened unprecedented opportunities for extraction of agricultural land use information. A series of promising research progress has been made. This review paper analyzes the state of the art of agricultural land use information extraction from four aspects:cropland, crop type, agricultural planting system, and agricultural land management. We found that: (1) The products of cropland mapping have been improved from the past coarse resolution (500~1000 m) to a higher spatial resolution of 10~30 m in the past decade. The global and regional cropland layers have been well established; but there is a need to track historical cropland changes, especially to identify the key turning points, by making full use of the existing remote sensing data (data fusion and satellite constellation approaches). (2) Existing crop type mapping efforts have been mostly carried out by combining ground survey data with satellite remote sensing (mainly Landsat and Sentinel-2). It has been operationalized in North America and Europe, but the ability to monitor crop planting areas needs to be strengthened in other countries including China. Also, the early season monitoring capacity of crop type mapping needs to be improved; (3) Existing studies on tracking agricultural planting systems are mainly concentrated in Eastern Europe (e.g., the abandonment after the breakup of the Soviet Union). In China, cropland abandonment, rotation, and fallow are also common in the recent decade, due to economic and policy factors; however, existing studies are lacking on this issue. (4) in terms of the agricultural land management, encouraging progress has been made on the regional products of irrigation, but the reliability and accuracy of the products need to be improved. New technologies, including the emerging multiple sources of remote sensing data so-called remote sensing big data, deep learning algorithms, and cloud computing platforms (e.g., Google Earth Engineand Amazon Web Services) provide unprecedented opportunities for future agricultural land use information extraction, which will rely on (1) the fusion of multi-source data to form remote sensing big data with higher spatial, spectral, and temporal resolutions, (2) coupling of intelligent methods such as machine learning and deep learning algorithms with expert knowledge-based methods considering geographical and phenological information, and (3) the application of cutting-edge technologies such as remote sensing cloud computing platforms.

Key words: agricultural land use, remote sensing, cropland mapping, crop type mapping, agricultural planting system, agricultural land management