【摘要】 为提高大范围作物遥感制图精度，利用Google Earth Engine(GEE)平台对Sentinel-2数据进行预处理，合成整个河北省2019年归一化植被指数（NDVI）、增强型植被指数（EVI）和二波段增强型植被指数（EVI2）的最大值影像；结合地理国情数据，输入自动统计分配算法以挑选出最优的植被指数；利用最优影像提取2016—2019年30 m分辨率的河北省冬小麦信息。结果表明，与验证样本相比，NDVI、EVI和EVI2的小麦分布用户精度均大于94%，制图精度均大于91%，总体精度均大于98%,Kappa系数均大于0.96，其中EVI的精度最高；与统计数据相比，所有小麦种植县2016—2019年的平均精度均大于97%。该自动统计分配算法无需依赖训练样本，能快速在其他年份复制使用，提高大范围作物分布制图效率，且获得的小麦信息与统计数据有良好的一致性，可为大范围小麦分布快速制图提供一定的技术方法参考和思路借鉴。更多还原

【Abstract】 In order to improve the accuracy of large-scale crop remote sensing mapping,the Sentinel-2 data was preprocessed by Google Earth Engine(GEE)platform,and the maximum images of NDVI,EVI and EVI2 of the whole Hebei Province in 2019 were synthesized. Combined with the data of geographical conditions,the automatic statistical distribution algorithm was input to select the optimal vegetation index. The optimal image was used to extract the winter wheat information of Hebei Province with a resolution of30 m from 2016 to 2019. The results showed that compared with the validation samples,the user accuracy of wheat distribution of NDVI,EVI and EVI2 was greater than 94%,the mapping accuracy was greater than 91%,the overall accuracy was greater than 98%,and the Kappa coefficient was greater than 0.96,of which EVI had the highest accuracy. Compared with the statistical data,the average precision of all wheat planting counties in 2016—2019 was greater than 97%. The automatic statistical allocation algorithm does not need to rely on training samples,and can be quickly copied and used in other years to improve the efficiency of large-scale crop distribution mapping,and the obtained wheat information has good consistency with the statistical data. It can provide some technical methods and ideas for rapid mapping of large-scale wheat distribution.更多还原