摘要
遥感技术已成为大宗作物种植面积提取的有效手段。为避免冬小麦提取中受光学数据缺乏的影响,基于随机森林算法(RF)和Google Earth Engine(GEE)云平台,探索时间序列Sentinel-1合成孔径雷达(SAR)数据后向散射系数对冬小麦提取效果,并融合Sentinel-1、2主被动遥感数据,研究后向散射系数、光谱特征、植被指数特征与纹理特征的不同组合对冬小麦识别精度的改善情况。结果表明:仅融合多时相Sentinel-1 SAR数据时,分类总体精度为85.93%,Kappa系数为0.75,冬小麦识别精度达到95%以上。融合多时相SAR数据与单时相光学数据,在充分利用极化信息和光谱信息进行分类后,分类总体精度为95.78%,Kappa系数为0.92,比多时相SAR分类结果分别提高9.85个百分点和约22.67%,对冬小麦的识别精度提高约2个百分点。通过分析不同特征组合情况下纹理特征对分类的影响,发现纹理特征对冬小麦的识别精度影响程度较小。
Remote sensing technology had become an effective method to extract planting area of bulk crop.With the aim to avoid the lack of optical data in winter wheat extraction,the validity of time series Sentinel-1 synthetic aperture radar(SAR)backscattering coefficients on winter wheat identification was explored based on random forest(RF)and Google Earth Engine(GEE)cloud platform.And Sentinel-1 and 2 active and passive remote sensing data was integrated to explore the improvement of winter wheat identification accuracy on combining various features groups of backscattering coefficients,spectral features,vegetation index features and texture features.The result indicated that the overall classification accuracy of the monthly average multi-temporal Sentinel-1 SAR polarization data was 85.93%,the Kappa coefficient was 0.75 and the winter wheat identification accuracy was above 95%.By integrating the monthly average time serious multi-temporal SAR data and the single-temporal optical data,the overall classification accuracy was 95.78%and the Kappa coefficient were 0.92.Integrating data fully used the polarization and spectral information and the overall classification accuracy and the Kappa coefficient were improved by 9.85 percentage points and 22.67%.The identification accuracy of winter wheat was improved by about 2 percentage points.The identification accuracy of winter wheat was affected by less than 0.9%by analyzing the influence of texture features under different features combinations.Therefore,the method and platform used accurately and efficiently obtained winter wheat planting area and it had a good promotion value.
作者
马战林
刘昌华
薛华柱
李静茹
房旭
周俊利
MA Zhanlin;LIU Changhua;XUE Huazhu;LI Jingru;FANG Xu;ZHOU Junli(School of Surveying and Land Information Engineering,Henan Polytechnic University,Jiaozuo 454003,China;Institute of Remote Sensing and Surveying and Mapping of Henan Province,Zhengzhou 450003,China)
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2021年第9期195-205,共11页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家自然科学基金项目(51779099)
国家自然科学基金面上项目(41721333)
河南省科技攻关重点项目(192102310270)
河南理工大学博士基金项目(B2017-09)。
