【作者】 蒋桂英；

【导师】 官春云； 陈金湘； 李少昆；

【作者基本信息】 湖南农业大学 ， 作物栽培学与耕作学， 2004， 博士

【摘要】 多光谱遥感以波长区域不连续的宽波段方式记录地物光谱，它以一个宽波段的单值代表一个相对宽光谱区域的辐射动态。与多光谱遥感相比，高光谱遥感可以将视域中观测到的各种地物以完整的光谱曲线记录下来，获取连续的光谱信息，光谱分辨率一般小于10nm，低于地表物质的诊断性光谱宽度（20-40nm）。其本身的波段数已达几百上千个，加上原始光谱的一阶导数、二阶导数、对数变换及其各种高光谱植被指数等，这并不是简单的数据量的增加，而是信息量的增加，信息量可增加十倍以至数百倍。鉴于高光谱遥感的独特性能，特别是在地表物质的识别和分类，有用信息的提取等方面的优势，使得这一技术在植被的精细分类、农作物的长势监测和农田水肥状况的分析等方面展现了巨大的应用前景。作物栽培生理指标的提取和产量估算的研究成为高光谱遥感技术最精华和核心的部分。 本文系统地研究了棉花主要栽培生理指标和产量与冠层光谱反射率之间的关系，采用两类分析方法，一是通过统计相关分析的方法建立观测光谱数据或由此衍生的植被指数与棉花栽培生理指标和产量之间的关系；二是基于棉花观测光谱特征变量的分析技术。以期①找出一些较适用于棉花栽培生理指标和产量估算的敏感光谱波段和植被指数。②建立高光谱特征变量估测棉花栽培生理指标和产量的模型。 研究内容包括：1)通过三年的不同水、肥和密度水平的棉花田间试验，以人为方式造成棉花长势的等级差异，作为本研究的基础；2)利用野外光谱仪在自然环境下测定不同长势、不同生育期的棉花冠层光谱，分析了棉花光谱在250-3500nm范围内的变化规律和棉花红边参数在各个生育期的变化规律；3)完成了叶绿素、叶片全氮、叶片可溶性糖、叶片及植株含水量、LAI、单株生物量和产量与350-2500nm冠层光谱反射率在各个生育期的统计相关分析，了解了各种生化组分在各个生育期的相关性较好的敏感波段；4)将光谱特征参量与各个生育期的栽培生理指标和产量进行了统计相关分析，挑选通用性的光谱参量，建立了各个生育期的栽培生理指标和产量的遥感模型。 本研究结果有以下几个方面： 1．分析了植被最典型的光谱特征，即红边光谱参数，随生育期的变化规律。红边位置在棉花现蕾以后以营养生长为主阶段，会向长波方向移动，即所谓的“红移”现象，而当进入开花以后以生殖生长为主的阶段，红边位置会向短波方向移动，出现“蓝移”现象;红谷置随生育期的变化与红边位置有相似的规律;红边宽度在棉花营养生长期是减小的，而从开花到吐絮阶段，红边宽度逐步增大。 2.完成了叶绿素、叶片全氮、叶片可溶性糖、叶片及植株含水量、LAI和单株生物量与350一25O0nm冠层光谱反射率在各个生育期的统计相关分析，分析了各种生化组分在各个生育期的相关性较好的敏感波段。从棉花盛蕾期至吐絮期各敏感波段中近红外波段770一85Onm处光谱反射率与LAI在全生育期都显著或极显著正相关;植株含水量与近红外770一950nm，1000一1300nln波段的光谱反射率在全生育期都有显著正相关特性，所以近红外波段770一850nln、770一95Onm，1000一130Onln处是棉花LAI和棉株含水量的指示性波段。 3.将各个生育期的栽培生理指标与冠层光谱特征参量进行了统计相关分析，挑选通用性光谱参量，建立了各个生育期的生化组分的遥感模型。绿峰（560nm）特征等是很好的色素组分反演参量，将叶绿素按3种表示方式进行处理，即干重含量、鲜重含量和叶绿素密度。3种数据与光谱反射率或光谱特征的相关性也不一致。通过比较分析，建立了基于67Onm吸收谷深度的叶绿素干重含量、叶绿素密度的分段遥感模型，和基于560nm绿峰反射峰特征参量的叶绿素鲜重含量模型。 从盛蕾期至吐絮期，利用棉花绿峰波段的反射峰高度，能够很好地反演叶片含水量，冠层光谱的67Onm处吸收谷深度与植株含水量始终都有很好的统计正相关关系，并建立了基于冠层光谱绿峰波段的反射峰高度和67Onm处吸收谷深度的叶片和植株含水量遥感反演模型。 对棉花重要的碳氮代谢指标，即叶片全氮含量、叶片可溶性糖含量等进行了研究，98Onm处弱吸收谷面积是估算叶片含氮量的敏感特征参量，抗大气植被指数VARI₇00是估算叶片可溶性糖的敏感特征参量，并建立了基于98Onm处弱吸收谷面积和抗大气植被指数vAR工₇00的各个生育期的叶片全氮含量、叶片可溶性糖含量的遥感模型。 LA工和生物量是监测作物长势好坏的重要的农学参数。LAI在各个生育期都有稳定的通用性光谱特征参量，即抗大气植被指数VAR工₇00;单株地上部分干生物量与67Onm处吸收谷的深度在盛蕾期至盛铃期呈极显著的正相关关系。并建立了基于670nln处吸收谷的深度和抗大气植被指数VARI₇00的单株生物量和LAI的遥感反演模型。 4.利用田间冠层光谱反射率和反射率光谱特征参量进行了棉花单位面积产量估算研究，发现抗大气植被指数VAR工₇00能够很好地预测棉花单位面积产量。 5.用高光谱遥感模型对开花结铃期棉株含水量以及LAI的估算效果较佳，其估算精度80%以上;其次是对叶绿素密度、单株生物量和叶片可溶性糖含量的估算，其估算精更多还原

【Abstract】 Multispectral remote sensing notes land spectra by non-continual wide wave band fashion in wave width region, which use one data of wide wave band represents radiation trend of a relative wide spectral region. By comparison with multispectral remote sensing, hyperspectra can completely note everything by a integrated spectral curve in its watching region and can receive continual spectral information, its spectral resolution is commonly less than 10nm and less than 20-40nm of diagnosis wide wave of land. It has several hundreds and thousands wave bands and first-order and second derivative spectra and every vegetation indices and so on. This is not increase in data number but is increase in information. The information capacity is add to several tenfold or hundreds; Because of the special characteristic of hyperspectra, especially identification and classification of land surface and extracting useful information and so on. This technique has tremendous applied foreground. The study on cultivating physiological characteristics and yield of crop is the most prime and core part of hyperspectra. The main objectives of the study includes:1) To find out some spectral wave bands and vegetation indices which would be applicable to estimate cultivating physiological characteristics and yield of cotton;2) The hyperspectral remote sensing models of different cultivating physiological characteristics and yield are built in different growth stages, which are based on the hyperspectral reflectance characteristic parameters.The main contents of the study are as following:1) This study is based on the rank difference of the nitrogenous nutrition level and watering level and density by man-made style through three years cotton farm experiment.2) Canopy spectral of cotton is measured by Analytical Spectral Devices (FieldSpec) in different growth stage. The study has analyzed the variation of cotton canopy reflectance with wavelength, compared the canopy reflectance curves of different nitrogenous nutrition level and watering level, also compared reflectance curves of leaf and cotton canopy. The study has also analyzed the variation of the spectral "red edge" features and meantime designed 46 hyperspectral reflectance characteristic parameters to build the hyperspectralremote sensing models of different cultivating physiological characteristics and yield in different growth stages.3) According to the statistical correlation coefficients between spectral reflectance in 350-2500nm range and cultivating physiological characteristics and yield are analyzed in all growth stages, all the sensitive bands for different cultivating physiological characteristics and yield are found.4) The hyperspectral remote sensing models of different cultivating physiological characteristics and yield are built in different growth stages, which are based on the 46 hyperspectral reflectance characteristic parameters.The research achievements are as following:1) The spectral "red edge" features are calculated and analyzed. Red-shift is found with early increase in cotton’s budding stage; Blue-shift is found with late decrease in cotton’s lint-opened stage. The red edge absorption width decreases in the early growth stages, and increases in the late growth stages.2) The statistical correlation coefficients between spectral reflectance in 350nm-2500nm range and cultivating physiological characteristics and yield are analyzed in all growth stages, so all the sensitive bands for different cultivating physiological characteristics and yield are found. The statistical correlation coefficients between LAI plant water content and spectral reflectance in near-infrared bands at 770-850nm 770-950nm, 1000-1300nm are significant in all growth stages.3) The statistical correlation coefficients between 46 hyperspectral reflectance characteristic parameters and cultivating physiological characteristics and yield are analyzed in all growth stages, the hyperspectral remote sensing models of different cultivating physiolog更多还原