【作者】 唐延林；

【导师】 王人潮； 黄敬峰；

【作者基本信息】 浙江大学 ， 土壤学， 2004， 博士

【摘要】 高光谱遥感能在特定光谱范围内以高光谱分辨率同时获取连续的地物光谱曲线，从而可以构成独特的超多维光谱空间，使得遥感应用着重于在光谱维上进行空间信息展开，获得目标物更多的光谱信息，为定量分析地球表面生物物理、生物化学过程提供参数和依据。高光谱遥感具有光谱分辨率高(一般＜10nm)、波段连续性强(在0.4～2.5μm范围内有几百个、上千个波段)、光谱信息量大等特点，其分辨率低于一般地表物质的半吸收宽度(约20～40nm)，因此，高光谱遥感所获得的数据和图像能以足够的光谱分辨率区分那些具有诊断性光谱特征的地表物质，使得高光谱遥感具有广阔的应用前景。 农作物的光谱特征是植被遥感的一个重要方面，它是农作物遥感识别、遥感长势监测和估产、遥感品质监测的重要依据。农作物光谱与农作物品种、植株密度、冠层结构、叶片形状、叶组织结构、农作物生化组分及比例、光谱测量条件(如气象条件、光谱仪分辨率、测量日期、背景)等因素有关。水稻高光谱特征及其理化基础和农学机理是水稻长势遥感监测、遥感估产和水稻品质遥感监测的主要依据。早期对水稻光谱的研究主要是针对遥感长势监测与估产，近年来对水稻光谱的研究主要集中在高光谱植被指数构建、导数光谱处理方法及辐射模型反演三个方面，目的是探索水稻光谱的物理化学基础和农学机理、进而提高水稻遥感长势监测和估产的精度，并利用水稻高光谱来估算其蛋白质、淀粉含量，进而探究水稻品质与其光谱特征之间的关系。 品种、生态环境、施肥量和施肥方法、栽种密度等因素都会影响水稻的株型结构、冠层结构、LAI、生物量、叶绿素密度、营养成分转移方法及速率等农学、生物化学参数，进而影响衡量水稻品质的两个重要参数——粗蛋白质含量和直链淀粉含量。而遥感理论和实践表明，植被株型结构、中文摘要冠层结构、LAI、生物量、叶绿素密度和生化组分及其含量的不同，其反射光谱大小及特征也会不同。因此，从理论上讲，通过监测水稻在各个不同发育时期(特别是抽穗后各时期)的光谱及其变化，可以预测稻米中的蛋白质和直链淀粉的积累，进而监测水稻品质。 本研究得到了国家“863”子课题“水稻品质遥感监测研究与应用”(2002AA243011)、国家自然科学基金项目“不同氮素水平的水稻高光谱遥感精确诊断机理与方法研究”(40 1 7 1 065)的资助，是在浙江大学农业遥感与信息技术应用研究所从1983年以来对水稻光谱研究的基础上，通过适当的水培和小区试验设计、测量与分析，获取不同品种在不同氮素水平下、不同发育时期水稻冠层和叶片的高光谱数据及水稻生育期间的平均气温等气象数据，测定水稻主要生化组分的高光谱、冠层和组分的生物物理与生物化学参数，对试验结果通过多种方法予以分析，建立水稻产量、主要生化组分的高光谱遥感估测模型，探索水稻品质遥感监测的可行性和初步建立水稻品质综合监测模型，进而探索水稻冠层、器官和生化组分的高光谱特征及其理化基础和农学机理。具体研究内容及结果如下: .通过两年水稻田间试验，培养出不同品种在不同氮素营养水平下的 水稻试验样本; .利用光谱仪系统测定不同品种水稻于不同氮素营养水平下在不同 时期的冠层、日一卜片、穗和稻谷光谱，及水稻主要生化组分(蛋白质、 粗淀粉、直链淀粉)的高光谱，比较了不同品种水稻冠层、叶片的 光谱差异、水稻不同生化组分的光谱特征及差异; .根据高光谱特点和水稻光谱特征，明确了水稻产量与冠层光谱高光 谱植被指数的相关性在抽穗前以RVI较好、在抽穗后以OVI较好， 并建立了水稻高光谱遥感估产模型，检验精度在95%以上; .采用单变量线性与非线性拟合模型和逐步回归分析，得到了单位面 积叶片全氮含量的最佳估测模型为原始光谱反射率的逐步回归模 型、单位面积土地上叶片全氮含量和单位面积土地上叶茎全氮含量 的最佳估测模型为一阶导数光谱值的逐步回归模型、稻穗粗蛋白 质、粗淀粉含量和稻谷粗蛋白质、粗淀粉含量的最佳估测模型为一 阶导数光谱值的逐步回归模型，模型检验精度在90%以上; .分析了不同品种之间、同一品种不同生长条件下时粗蛋白质、淀粉 和直链淀粉含量差异，建立了水稻品质两个重要指标—粗蛋白质中文摘要 尸和直链淀粉A含量的一般预测模型，可以根据水稻的冠层光谱和 灌浆期的日平均温度来预测， P(A)=S(兄)*(kT+b) 模型检验精度一般在90%以上; .分析了水稻冠层和室外廿卜片的红边特征，明确了它们的红边“双峰” 和“多峰”现象并不是由导数光谱的计算方法引起的，而可能是由 光源、气象条件、冠层结构等因素引起的; .对水稻蛋白质和粗淀粉的混合光谱分析发现，混合物的反射光谱与 纯净物反射光谱相比会出现峰、谷位置“红移”或“蓝移”现象， 且稻米蛋白质和粗淀粉含量与其光谱曲线在2020一2235nm之间的 吸收面积S具有显著相关。 本研究的技术新意和创新点有以下三个方面:.本研更多还原

【Abstract】 Hyperspectral remote sensing can note the continual spectral curves of ground objects with hyperspectral resolution at the same time in specifical waveband range. They may form particular super-multidimensional spectral space to bring the application of remote sensing emphasize on space information on spectral dimension for obtaining more spectral information. This will provide parameters and basis for quantitive analysis of the biophysical and biochemical processes on the ground surface. It has high spectral resolution (generally less than 10nm), strong waveband continuity (several hundreds and thousands wavebands from 0.4 u m to 2.5 um and a great deal of spectral information. Its spectral resolution is commonly less than the half absoring width (about 20~40nm) of common objects. Therefore the data and pictures obtained by hyperspectral remote sensing are callable of distinguishing objects having diagnostic characteristics with adequate spectral resolution. It has wide application foreground.The spectral characteristics of crops is an significant aspect of remote sensing of vegetation and an important basis of distinguishing crops, monitoring their growth, estimating their yield and quality by remote sensing. The crops spectra are relate to their variety, density, canopy structure, leaf shape, leaf structure, biochemical components and rate of all biochemical components and conditions of measuring spectrum such as meteorology, resolution of spectroradiometer, measuring date and background. The hyperspectral characteristics of rice and its physical and chemical basis and agronomic mechanism are the primary foundation of its growth monitoring, yield estimating and quality monitoring by remote sensing. The study of rice spectrum mostly aimed at its growth monitoring and yield estimating in early time. This study mostly focused on constitution hyperspectral vegetationindex, method of derivative spectrum and retro-deduction of radiant models. The purpose is to search after the physical and chemical basis and agronomic mechanism of rice spectrum, improve the precision of growth monitoring and yield estimating, and estimating the contents of its crude protein and amylose. The final aim is at exploring the relation between rice quality and its spectral characteristics.The factors such variety, entrionment, quantity and method of fertilization and density could influence on the agronomic and biochemical parameters of rice such as its structure of plant and canopy, leaf area index LAI, biomass, chlorophyll density and transfer method and speed of nutrition. They cuould more influence on two important parameters weighing rice quality such as the contents of crude protein and amylose. The theory and practice of remote sensing indicate that vegetation with different structure of plant and canopy, leaf area index LAI, biomass, chlorophyll density, biochemical component and their contents will bring on different value and characteristics of reflective spectra. It may theoretically predict the accumulation of crude protein and amylose of rice by monitoring the rice spectrum and its change at different stages especially after heading. In this way, the quality of rice could be monitored.This study is Supported by the National ’863’ Project of China "Study on Monitoring Quality of Rice by Remote Sensing and Its Application" (2002AA243011) and the National Natural Science Foundation of China " Study on Mechanism and Method of Precisely Diagnose Hyperspectra of Rice under Different Nitrogen Levels" (40171065). On the basis of rice spectrum having been studied by Institute of Agricultural remote sensing & information application from 1983, it is to obtain the hyperspectral data of canopy and leaf of different variety of rice at different stages under different nitrogen levels and some meteorologic data such as average temperature in growth stages, determine the hyperspectral data of main biochemical component and the biophysical and biochemical parameters of canopy and organ. It is to establish estimating models of yi更多还原