基于时序Sentinel-2影像的梨树县作物种植结构  

Crop planting structure in Lishu County based on time-series Sentinel-2 images

作　　者：刘俊伟 陈鹏飞[1,3] 张东彦 赵红伟[4] LIU Jun-wei;CHEN Peng-fei;ZHANG Dong-yan;ZHAO Hong-wei(Institute of Geographic Sciences and Natural Resources Research,Chinese Academy of Sciences/State Key Laboratory of Resources and Environmental Information System,Beijing 100101,China;National Engineering Research Center for Agro-Ecological Big Data Analysis&Application,Anhui University,Hefei 230601,China;Jiangsu Center for Collaborative Innovation in Geographical Information Resource Development and Application,Nanjing 210023,China;Institute of Agricultural Resources and Regional Planning,Chinese Academy of Agricultural Sciences,Beijing 100081,China)

机构地区：[1]中国科学院地理科学与资源研究所/资源与环境信息系统国家重点实验室,北京100101 [2]安徽大学农业生态大数据分析与应用技术国家地方联合工程技术研究中心,安徽合肥230601 [3]江苏省地理信息资源开发与利用协同创新中心,江苏南京210023 [4]中国农业科学院农业资源与农业区划研究所,北京100081

出　　处：《江苏农业学报》2020年第6期1428-1436,共9页Jiangsu Journal of Agricultural Sciences

基　　金：国家重点研发计划项目(2017YFD0201501);国家自然科学基金项目(41871344);中国科学院战略先导科技专项(XDA23100100)。

摘　　要：Sentinel-2影像具有空间分辨率高,重访周期短的优势。本研究探讨了基于Sentinel-2数据开展梨树县作物精准分类,进而进行种植结构分析的可行性。为此,收集了多时相的Sentinel-2影像和大量地面不同作物类型样点数据。基于以上数据,选择决策树法、最大似然法、支持向量机法等3种经典分类方法开展影像分类,并对它们的结果进行对比以选择最优分类方法。然后,利用最优分类方法获得的分类结果对梨树县作物种植结构进行分析。结果表明,基于时序Sentinel-2影像,利用作物的物候特征与其光谱特征之间的联系可实现对梨树县作物种植面积和空间分布信息的准确提取,从而对该县种植结构进行客观评价。3种分类方法中,决策树法的分类精度最高,其总体分类精度为93.53%,Kappa系数达到0.8906。Sentinel-2 images have the advantages of high spatial resolution and short revisit cycle.This study investigated the feasibility of making accurate classification of crops and planting structure analysis in Lishu County based on Sentinel-2 data.Multi-temporal Sentinel-2 images and a large number of ground sample data of different crops were collected.Based on the above data,three classic classification methods including decision tree method,maximum likelihood method and support Vector machine method were selected in image classification,and their results were compared to select the best classification method.Then,the classification results obtained by the best method were used to analyze the crop planting structure of Lishu County.The results showed that based on time-series Sentinel-2 images,the planting area and spatial distribution information of crops in Lishu County could be extracted accurately by combining the phenological characteristics and spectral characteristics of the crops,then the planting structure of the county could be evaluated objectively.Among the three classification methods,the decision tree method got the best results,with the overall classification accuracy of 93.53%and the Kappa coefficient value of 0.8906.

关 键 词：作物种植结构 Sentinel-2 光谱特征 

分 类 号：S127[农业科学—农业基础科学]