基于作物参考曲线法的冬小麦NDVI时间序列重建——以河北省部分区域为例  

NDVI Time Series Reconstruction of Winter Wheat based on Crop Reference Curve Method:a Case Study of Some Areas of Hebei Province

作　　者：敖洋钎 王永前[1] 孙政 孙亮[2] AO Yangqian;WANG Yongqian;SUN Zhen;Sun Liang(College of Resources and Environment,Chengdu University of Information Technology,Chengdu 610225,China;Institute of Agricultural Resources and Regional Planning,Chinese Academy of Agricultural Sciences,Beijing 100081,China)

机构地区：[1]成都信息工程大学资源环境学院,四川成都610225 [2]中国农业科学院农业资源与农业区划研究所,北京100081

出　　处：《成都信息工程大学学报》2025年第2期238-244,共7页Journal of Chengdu University of Information Technology

基　　金：国家重点研发计划资助项目(2022YFD2001102)。

摘　　要：近年来,多源传感器融合的遥感技术发展快速,弥补了单一数据源存在的时间分辨率或空间分辨率难以达到研究要求的情况。然而这种方法受到融合影像质量的影响较大,在部分区域难以完成对研究区作物高时空分辨率的归一化植被指数(NDVI)的计算。本文采用从冬小麦分布图中识别MODIS冬小麦像元并提取多组NDVI时间序列作为作物参考曲线,并利用Sentinel-2数据对作物参考曲线进行挑选以及调整从而实现重建的方法,但该研究方法受到作物参考曲线的影响。因此在重建的基础上进一步测试作物参考曲线算法在使用不同年份参考曲线的情况下的重建精度。精度评价结果表明,在作物快速生长的时期重建NDVI的相对误差(RE)在10%左右,在缓慢生长的NDVI高值时期重建的RE则在4%~7%且生长期间中重建NDVI的均方根误差(RMSE)均低于0.17。因此,重建后的每日30m分辨率NDVI图具有较好的质量可以满足复杂农业场景下提供高质量NDVI数据的需求。In recent years,the remote sensing technology of multi-source sensor fusion has been developed rapidly,which makes up for the fact that the temporal resolution or spatial resolution of a single data source is dfficult to meet the research requirements.However,this method is greatly affected by the quality of fused images,so it is difficult to complete the calculation of the normalized vegetation index(NDVI)with high spatiotemporal resolution of crops in the study area in some areas.In this paper,we used the method of identifying MODIS winter wheat pixels from the distribution map of winter wheat,extracting multiple sets of NDVI time series as crop reference curves,and using Sentinel-2 data to select and adjust the crop reference curves to achieve reconstruction,but the research method was affected by the crop reference curves.Therefore,on the basis of the reconstruction,the reconstruction accuracy of the crop reference curve algorithm was further tested when the reference curves of different years were used.The results showed that the relative error(RE)of reconstructed NDVI was about 10%during the period of rapid crop growth,and about 4%to 7%during the period of high value of NDVI of slow growth,and the root mean square error(RMSE)of reconstructed NDVI during the growth period was less than 0.17.Therefore,the reconstructed daily 30 m resolution NDVI map has good quality and can meet the needs of providing high-quality NDVI data in complex agricultural scenarios.

关 键 词：作物参考曲线 冬小麦 NDVI时间序列 Sentinel-2 MODIS 河北省 

分 类 号：S162.54[农业科学—农业气象学]