春玉米LAI和叶片氮素营养及产量的高光谱估测模型研究

发布时间：2018-03-07 06:36

  本文选题：高光谱　切入点：春玉米　出处：《内蒙古农业大学》2016年博士论文　论文类型：学位论文

【摘要】：高光谱技术即为精确农业的一种重要技术手段,因其具有方便、快捷、高效、对植株无损害等优点,己被大量应用于作物生长监测、作物植株水分监测、营养状况监测、作物产量评估、品质监测等多个方面。本研究以玉米为研究主体,通过分析不同种植条件下(不同品种,不同密度,不同施氮量,氮、密互作)玉米冠层及叶片高光谱特征与其相对应的LAI、SPAD值、叶片氮含量和产量的响应规律,明确4个理化指标的敏感波段,并利用光谱指数NDVI、RVI、DVI构建了基于高光谱植被指数的LAI、SPAD值、叶片氮含量和产量高光谱估测模型。其主要研究结果如下：(1)栽培环境的改变,会直接引起玉米生理生态参数变化,而这种变化又会因其栽培措施的不同而产生差异,例如：生育期(叶片衰老)、种植密度、施氮量等因素均会导致玉米LAI发生改变,但种植密度对LAI的影响最大,叶片衰老变化次之,施氮量最小。同理,不同栽培条件也会影响冠层和叶片高光谱特征,在可见光350-760nm波段,冠层光谱与叶片光谱反射率随着生育期进程呈增大趋势；不同种植密度下,冠层光谱反射率表现为随密度的增大而减小,而叶片光谱反射率则表现为随密度增加呈增大趋势；不同施氮量下,冠层光谱与叶片光谱反射率随着施氮量的增加呈下降趋势。在780-1300rnm波段,冠层光谱与叶片光谱反射率随生育进程呈逐渐下降趋势；不同种植密度下,冠层光谱反射率随密度的增大呈增大趋势,而叶片光谱反射率则无明显变化规律；不同施氮量下,冠层光谱反射率随施氮量的增加呈增大趋势,而叶片光谱反射率在这一波段则无明显变化规律。种植密度对冠层光谱反射率的影响大于施氮量,施氮量对叶片光谱的影响要大于密度。(2)通过对不同栽培条件下的玉米冠层、叶片光谱与LAI、叶片SPAD值、LNC和产量的相关分析,得出4个指标的反射率光谱敏感波长主要位于550nm、678nm、710nm和1100nm附近,一阶导数光谱敏感波长位于500nm、550nm、580-680nm之间、700nm和755nm附近。利用NDVI、RVI和DVI三种光谱参数构建了不同栽培条件下的LAI、SPAD值、LNC和产量高光谱估测模型,并对模型应用精度进行了比较,得出：模型在应用于其他栽培条件时均会出现较大偏差,其中冠层光谱模型对其他条件下各指标的估测精度都比较差,尤其是对LAI的估测偏差最大。而SPAD值和LNC的叶片光谱模型,具有较高的普适性。(3)不同栽培条件下高光谱参数值与各指标数值之间定量关系的差异是导致各指标高光谱估测模型普适性差的根本原因,对不同栽培条件下的光谱与各指标数据进行综合分析,可以降低两者之间的不匹配程度,提高模型的普适性。同时,利用土壤线参数和叶片光谱与冠层光谱反射率差值参数可以提高光谱模型的稳定性。在所建高光谱模型中,通用性较好的各指标估测模型有：LAI为NDVI(729.3,963.6)和MNDVI(729.3,963.3)模型；SPAD值和LNC为NDVI(729.3,963.6)和mRVI(729.3,963.6)模型；产量为NDVI(R'695.7, R'755.5)和MRVI(550.2,963.6)模型。
[Abstract]:Hyperspectral technology is an important means of precision agriculture, because of its convenient, fast, efficient, has the advantages of no damage to the plant, has been widely used in crop growth monitoring, crop water monitoring, status monitoring, crop yield assessment, many aspects of quality monitoring. In this study, corn the research subject, through the analysis of different planting conditions (different varieties, different densities, different nitrogen, nitrogen, close interaction) characteristics of corn canopy and leaf hyperspectral and the corresponding LAI, SPAD value, nitrogen content and leaf yield response rules, clear 4 physicochemical indexes and the sensitive bands. Using the spectral index NDVI, RVI, DVI constructed Hyperspectral Vegetation Index Based on LAI, SPAD value, nitrogen content and yield of leaf Hyperspectral Estimation Models. The main results are as follows: (1) the cultivation of the environment changes, may be a direct cause of maize physiological and ecological parameters But, this change will have the difference, because of the different cultivation measures such as growth period (leaf senescence), planting density, nitrogen and other factors will lead to the change of maize LAI, but the effects of planting density on LAI maximum, leaf senescence of nitrogen is minimal. Similarly, different cultivation conditions will also affect the spectral characteristics of canopy and leaf, in the visible band 350-760nm, canopy spectral reflectance and leaf growth process with increased; under different planting density, canopy spectral reflectance was observed with the increase of density decreases, while the leaf spectral reflectance is increased with the increase of density; under different nitrogen levels, canopy spectral reflectance and leaf spectral reflectance with the increase of nitrogen decreased. In 780-1300rnm band, canopy spectral reflectance and leaf spectral reflectance decreased gradually with the growth and development trend Potential; under different planting density, canopy spectral reflectance with increasing density increased, while the leaf spectral reflectance had no significant variation; under different nitrogen levels, the canopy spectral reflectance with the increase of nitrogen increased, while the leaf spectral reflectance in the wave period had no significant variation effect of planting density. The canopy reflectance is greater than the amount of nitrogen, the nitrogen effect on leaf spectra than density. (2) through the canopy of Maize under different cultivation conditions, leaf spectra and leaf SPAD values, LAI, LNC correlation analysis and yield, reflectance spectrum sensitive wavelength draw 4 index are mainly located at 550nm. 678nm, near 710nm and 1100nm, between the first derivative spectra are sensitive to wavelengths of 500nm, 550nm, 580-680nm, 700nm and 755nm. Near by NDVI, RVI and DVI three spectral parameters to build different cultivation conditions of LAI, SPAD The value of LNC, and the yield of Hyperspectral Estimation Models, and the application of the precision of the model were compared to that in the model applied to other cultivation conditions will appear larger deviation, the estimation precision of canopy spectral model for each index under other conditions are relatively poor, especially on the estimation of maximum deviation of LAI and SPAD value. LNC and leaf spectral model, general higher. (3) the difference of spectral parameters under different cultivation conditions and the quantitative relationship between the index value of each index is the result of numerical Hyperspectral Estimation Model of root causes of poor universality, data spectrum of different cultivation conditions and various indicators analysis, can reduce the degree of mismatch between the two, to improve the universality of the model. At the same time, the use of soil line parameters and leaf reflectance and canopy reflectance difference parameters can improve the stability of the model. In the spectrum In the hyperspectral model, the index estimation models with good universality are: LAI is NDVI (729.3963.6) and MNDVI (729.3963.3) model; SPAD value and LNC are NDVI (729.3963.6) and mRVI (729.3963.6) model; output is the model of "729.3963.6" and "X".

【学位授予单位】：内蒙古农业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：S513
，

本文编号：1578346