棉花全生长周期机载高光谱正射影像数据集构建

Construction and analysis of airborne hyperspectral orthophoto data set in the whole growth cycle of cotton

棉花作为我国的主要经济作物,对其开展低空遥感精细监测的研究目前尚缺乏高质量的监测样本数据集。高光谱影像具有丰富的光谱特征和空间特征,被广泛用于生化参量反演、病虫害监测、长势评估与产量预测。为此,本文构建了面向棉花全生长周期的机载高光谱成像数据集(Airborne hyperspectral imaging dataset for the full growth cycle of cotton,AHS-FGCC)。基于大疆M600 Pro无人机搭载Rikola高光谱成像仪,以100 m航高获取同一区域同一棉花品种苗期、苗后期、蕾期、花期、铃期、盛铃期以及吐絮期共7期无人机高光谱数据。该数据经过标准预处理,参考《地理信息元数据》规范,补充无人机载高光谱棉花数据元数据信息,从而构建全生长周期棉花高光谱正射影像数据集。通过与采样点ASD(SR-3500)地物光谱仪棉花冠层光谱曲线对比,分析棉花数据集的光谱精度以及对应的棉花生长期光谱变化规律。结果表明Rikola成像光谱仪和ASD地物光谱仪波长在503~850 nm区间,反射率曲线趋势具有良好的一致性,“绿峰特征”,“红谷特征”,“红边特征”3种典型光谱特征表现也基本吻合。本数据集可以较好地反应棉花不同生长周期的光谱特征变化规律,可为棉花的低空遥感精细监测提供样本数据,并为农作物高光谱或多光谱数据集构建的相关研究提供参考和借鉴。(数据集发布于科学数据银行,访问地址:https://www.scidb.cn/detail?dataSetId=a07c3451e3cf4ff090b1cbb72edc08c3。)

Cotton is the main economic crop in China,but there is still a lack of high-quality monitoring sample datasets for research on low-altitude remote sensing fine monitoring.Hyperspectral images not only have rich spectral and spatial characteristics,but also are widely used in biochemical parameter inversion,plant diseases and insect pests monitoring,growth potential assessment,and yield prediction.To this end,this paper builds an airborne hyperspectral image dataset for the full growth cycle of cotton(AHS-FGCC).Based on the DJI M600 Pro drone equipped with a Rikola hyperspectral imager,a total of 7 phases(seedling,seedling-seeding,bud,florescence,boll-forming,peak bolling and open-boll periods) of UAV hyperspectral data of the same cotton varieties in the same area with a height of 100 m.This data has been standardly preprocessed.Referring to the Geographic Information Metadata specification,the airborne hyperspectral cotton metadata information is supplemented to build a full growth cycle cotton hyperspectral orthography image dataset.The results show that the reflectance curves obtained by the Rikola imaging spectrometer and the ASD ground object spectrometer have good consistency in the wavelength range of 503-850 nm.In addition,the three typical spectral features of "green peak feature","red valley feature" and "red edge feature" are basically consistent.This dataset can better reflect the spectral characteristics of cotton in different growth periods and can provide sample data for fine monitoring of cotton by low-altitude remote sensing,and provide a reference for related research on the construction of crop hyperspectral or multispectral datasets.(Dataset published on Science Data Bank.Address:https://www.scidb.cn/detail?dataSetId=a07c3451e3cf4ff090b1cbb72edc08c3)