基于Hyperion高光谱影像的冬小麦地上干生物量反演

在黄淮海粮食主产区选择河北省衡水市深州市为试验区,以冬小麦地上干生物量为研究对象,以作物冠层高光谱和EO-1 Hyperion高光谱卫星数据为主要数据源,在分析冠层高光谱构建的窄波段植被指数(N-VIs)与实测冬小麦地上干生物量间相关性基础上,提出了利用拟合精度R2极大值区域重心确定冬小麦干生物量敏感的光谱波段中心的方法,并运用该方法确定了冬小麦生物量敏感波段中心。在此基础上,以敏感波段中心筛选结果为指导,利用窄波段植被指数及相关波段开展Hyperion高光谱卫星遥感区域冬小麦干生物量遥感反演和精度验证。最终,按精度最高原则优选区域冬小麦地上生物量反演结果。其中,研究采用了冬小麦孕穗期Hyperion数据,涉及的植被指数包括窄波段归一化植被指数(N-NDVI)、窄波段差值植被指数(N-DVI)和窄波段比值植被指数(N-RVI)。结果表明,通过与实测冬小麦地上干生物量对比,利用冠层高光谱冬小麦地上干生物量反演敏感波段筛选结果及其相应波段构建的Hyperion窄波段植被指数进行孕穗期作物干生物量估算取得了较好结果,其精度由大到小为:NNDVI、N-RVI、N-DVI。其中,以波段B18(波长528.57 nm)、波段B82(波长962.91 nm)构建的Hyperion N-NDVI估算区域冬小麦地上干生物量精度最高,相对误差(RE)和归一化均方根误差(NRMSE)分别为12.65%和13.78%。

Retrieving Winter Wheat Above-ground Dry Biomass Based on Hyperion Hyperspectral Imagery

How to efficiently use hyperspectral remote sensing data to quantitatively retrieve bio-physical and bio-chemical crop parameters and accurately obtain regional crop growth information has always been one of the hot issues in agricultural quantitative remote sensing researches. Based on optimal selection of sensitive hyperspectral bands, the research on retrieval of winter wheat above-ground dry biomass (ADBM) from Hyperion hyperspectral imagery was carried out in Shenzhou County, Hebei Province of Huang-Huai-Hai Plain which was the major grain producing region in China. Firstly, through analyzing relationship and their coefficient of determination (R2) values between field-measured ADBM of winter wheat and narrow band vegetation index (N-VIs) from crop canopy hyperspectral data, the method of determining hyperspectral sensitive band centers based on areas weight of R2 maximum values was proposed and applied. Then, supported by results of hyperspectral sensitive band centers, Hyperion hyperspectral remote sensing data was used to retrieve winter wheat ADBM at regional scale by using the N-VIs, and the accuracy of winter wheat biomass estimation results was validated. The Hyperion remote sensing data was acquired on Apr. 23, 2014, which was at the booting stage of winter wheat, and the used N-VIs included narrow band normalized difference vegetation index (N-NDVI), narrow band difference vegetation index (N-DVI) and narrow band ratio vegetation index (N-RVI). Compared with field-measured winter wheat ADBM, based on optimal selection of sensitive hyperspectral bands and Hyperion N-VIs constructed by the selected sensitive bands, the method of using Hyperion N-VIs to retrieve winter wheat ADBM had better performance, and the accuracy order of winter wheat ADBM of the N-VIs were determined, showing a descending trend as follows: N-NDVI, N-RVI and N-DVI. Among them, based on the Hyperion N-NDVI constructed by the selected sensitive bands (528.57nm and 962.91nm), the retrieval result of ADBM of winter wheat was the best and the relative error (RE) and normalized root mean square error (NRMSE) were 12.65% and 13.78%, respectively. It was proved that based on optimal selection of sensitive hyperspectral bands, using Hyperion N-VIs to retrieve winter wheat ADBM had certain feasibility and effectiveness. It could provide a new thought thread for hyperspectral remote sensing sensitive bands selection and for the improvement of quantitatively retrieving bio-physical and bio-chemical crop parameters.