万丛 孙智虎 梁治华 张锦水

摘要 2019年11月3日发射的GF-7号卫星是我国的第二颗亚米级、多角度民用商业卫星，其在农作物面积分布精细化识别方面潜力有待评估。依据2018年国家统计局数据，全国冬小麦播种面积占粮食作物总播种面积的19.23%，通过遥感手段准确识别冬小麦分布情况，是作物长势和作物估产等后续遥感产品准确评估的保证，对确保粮食安全具有极其重要的意义。通过支撑向量机和随机森林2种机器学习算法，分析高分七号亚米级光谱特征及其纹理特征对冬小麦的精细化识别能力。结果表明，基于影像光谱特征，SVM分类器取得了最优的分类精度，其中冬小麦识别精度为93.96%，总体精度为91.01%，Kappa系数为0.763 2，面积精度为91.46%。

关键词 GF-7;遥感;冬小麦;随机森林;支撑向量机

中图分类号 S-127  文献标识码 A

文章编号 0517-6611（2021）12-0244-04

doi：10.3969/j.issn.0517-6611.2021.12.064

开放科学（资源服务）标识码（OSID）：

Refined Identification of Winter Wheat Area Based on GF-7 Satellite Remote Sensing

WAN Cong1，SUN Zhi hu2，LIANG Zhi hua3  et al （1.Data Management Center，National Bureau of Statistics of the Peoples Republic of China，Beijing 100826;2.China University of Geosciences （Beijing），Beijing 100083;3.Institute of Remote Sensing Science and Engineering，Faculty of Geographical Science，Beijing Normal University，Beijing 100875）

Abstract GF 7 satellite launched on November 3，2019，is Chinas second sub meter multi angle commercial satellite. Its potential in fine identification of crop area distribution needs to be evaluated.According to the data of the National Bureau of Statistics in 2018，the sown area of winter wheat accounts for 19.23% of the total sown area of grain crops.Accurate identification of winter wheat distribution by remote sensing means is the guarantee for accurate evaluation of crop growth and crop yield estimation and other follow up remote sensing products，which is of great significance to ensure food security.In this research，support vector machine （SVM） and random forest machine learning algorithms were used to analyze the fine recognition ability of the spectral features and texture features of GF 7 images.The results showed that the SVM classifier achieves the optimal classification accuracy based on the spectral characteristics of the image.The recognition accuracy of winter wheat was 93.96%，the overall accuracy was 91.01%，the kappa coefficient was 0.763 2，and the area accuracy was 91.46%.

Key words GF 7;Remote sensing;Winter wheat;Random forest;SVM

近年來，国产高分辨率遥感卫星的相继发射，尤其是亚米级分辨率卫星，使得影像上可以观测到更为细致的地表覆盖结构。虽然中国农业种植结构在朝着规模化种植的趋势发展，但在广大农村小农经济下作物种植结构比重仍不可忽视，其农业生产具有超小规模化和农地细碎化等特征[1]。中、低分辨率遥感影像难以准确识别农作物种植破碎情况，新发射的GF-7亚米级卫星影像为精细化识别农地种植分布提供了可能。

由于遥感数据能反映真实的地物类型特点，越来越被广泛用于农业调查，尤其是第三次农业普查首次采用遥感数据应用于农作物面积普查中[2]，通过遥感进行地物测量替代了大量的人工实地测量工作。……