干旱区地理2024，Vol.47Issue(4)：672-683,12.DOI:10.12118/j.issn.1000-6060.2023.262
基于Sentinel-2时序数据的新疆焉耆盆地农作物遥感识别与评估
Remote sensing identification and assessment of crops in the Yanqi Basin,Xinjiang,China based on Sentinel-2 time series data
摘要
Abstract
To obtain timely and accurate information about crop cultivation in arid zones,this study used the PIE-Engine Studio platform to extract 14 vegetation indices in the Yanqi Basin,Xinjiang,China based on 2022 Senti-nel-2 images and 1948 field location sampling data during the crop reproduction period.Crop planting informa-tion was extracted using the See5.0 decision tree,random forest(RF),and multiple regression(MR)models to se-lect feature parameters.Each model was combined with support vector machine(SVM)algorithms to construct five classification models and five sample segmentation schemes.The best classification scheme was determined by visual interpretation and confusion matrix comparison.The results are as follows:(1)The overall accuracy(OA)and Kappa coefficients of all classification models are above 92.20％and 0.9037,respectively,indicating that it is feasible to extract crop information using the SVM algorithm in the PIE platform.(2)The mean OA and Kappa coefficients of SVM-with-red-edge are 93.77％and 0.9236,which are 0.96％and 0.0120,respectively.(3)The introduction of vegetation index improved the OA and Kappa coefficients of SVM-RF,SVM-MR,and SVM-See5.0 compared with the SVM-with-red-edge method.(4)The mean OA and Kappa coefficient relationships for the five classification models were SVM-RF>SVM-MR>SVM-See5.0>SVM-with-red-edge>SVM-without-red-edge,showing that the inclusion of the red-edge band and vegetation index significantly improved the accuracy of crop identification,with SVM-RF(8:2)being the best classification model with OA and Kappa coefficients of 98.72％and 0.9866,respectively.These results provide new ideas and references for accurate and rapid access to large-scale arid zone crop information.关键词
农作物/Sentinel-2/支持向量机/PIE-Engine Studio/焉耆盆地Key words
crop/Sentinel-2/support vector machines/PIE-Engine Studio/Yanqi Basin引用本文复制引用
张旭辉,玉素甫江·如素力,仇忠丽,亚夏尔·艾斯克尔,阿卜杜热合曼·吾斯曼..基于Sentinel-2时序数据的新疆焉耆盆地农作物遥感识别与评估[J].干旱区地理,2024,47(4):672-683,12.基金项目
国家自然科学基金项目(U1703341,41764003) （U1703341,41764003）
自治区科技创新基地建设计划项目(2020D04039)资助 （2020D04039）
