【作者】 吴琼；

【导师】 王纪华；

【作者基本信息】 吉林大学 ， 农业电气化与自动化， 2012， 博士

【摘要】 小麦幼苗的发育好坏直接影响最终的产量、品质及安全性，苗期长势是农艺学家、育种家、生物化学家非常关注的作物特性。目前，作物长势无损监测研究大多侧重于采用光谱技术获取中后期的群体参数（叶面积、生物量、覆盖度等）实现。而幼苗的光谱检测研究还较少，除了限于个体分析，在大范围无损监测方法方面更缺乏。光谱成像技术具有“图谱合一”的特性，它能同时获取观测对象的光谱与图像信息，且有很高的光谱分辨率和更精细的波段，海量的图谱信息在研究作物个体差异性方面具有得天独厚的优势。论文利用高光谱成像仪，开展了盐、冻害、干旱等不同胁迫类型下的苗期小麦长势监测研究及穗发芽程度识别研究，初步探索了基于高光谱成像技术的小麦苗期长势监测方法，这为地面、航空和卫星遥感监测小麦苗情提供了理论依据。研究的具体内容如下：1.利用高光谱成像诊断小麦苗期盐胁迫程度。通过对中国春（对盐敏感）、洲元9369（普通高产）、长武134（耐盐）的小麦幼苗进行盐胁迫处理，活体采集126个小麦叶片样本在400～1000nm波段范围内的成像光谱数据，分析小麦幼苗叶片光谱与相对叶绿素含量值（SPAD）间的相关性，利用成像光谱仪图谱合一的优势，对盐胁迫状态下小麦幼苗的冠层、单株、叶片的不同尺度、不同区域的特征进行分析，对其叶绿素组分分布情况进行诊断。结果表明在693nm处相关系数绝对值最大，筛选为小麦幼苗的特征波长，通过建立回归模型，进行叶绿素反演填图，从填图结果中可以直观看出叶片不同部位叶绿素的分布情况。为了提高模型预测效果，更充分利用光谱信息，采用偏最小二乘法（PLS）建立预测模型，其相关系数R=0.8138，标准差SEC=4.7552。表明成像光谱能够预测不同盐胁迫处理下的小麦幼苗的组分信息，从而根据组分含量高低对盐胁迫程度进行诊断。2.利用高光谱成像监测小麦苗期冻害胁迫。2010年冬季在北京种植30个南方地区主栽的抗冻性较差的小麦品种，在三个不同时期进行冻害程度检测。采集小麦叶片样本在450～900nm波段范围内的成像光谱，并利用多酚测定仪测量小麦的氮平衡指数。结果表明，随着冻害程度的加深，在450～650nm波段，光谱的反射率逐渐增加，而在700～900nm波段，光谱反射率逐渐减小。通过提取光谱图像中小麦的平均光谱，对小麦幼苗冠层的光谱反射率与氮平衡指数进行相关性分析，最大负相关系数出现在650nm～700nm之间的红边位置，相关系数约为-0.7。在680nm处，三个时期冻害胁迫下的相关系数绝对值均为最大，筛选为小麦幼苗冻害诊断的特征波长，根据特征波长下的光谱图像，可直观看出小麦受冻害的区域及程度。研究表明基于高光谱成像技术监测小麦幼苗冻害胁迫是可行的，能够精确反映小麦幼苗受冻害的部位。3.利用高光谱成像监测小麦苗期干旱胁迫。分别在三个时期采集小麦幼苗发生干旱胁迫时的成像光谱数据，并记录相应的苗期单株叶片的SPAD值，发现随着干旱胁迫程度的加重，扬麦15、扬麦158、京冬8三个抗旱性较差的小麦中，测得的单株叶片的SPAD平均值逐渐减小，而抗旱性较好的农大211所测得的叶片SPAD值在两个时期的测量值相近，差异不大。分析发现在450nm～700nm光谱波长范围内，小麦干旱胁迫最严重时，其冠层平均光谱反射率最高，且不具有明显的绿色植被的光谱反射率特征波峰。通过对苗期干旱胁迫下的冠层高光谱反射率进行一阶微分，并与所测的叶绿素SPAD平均值建立相关系数曲线，发现在700～780nm范围内光谱与SPAD值正相关，表明干旱胁迫可能对光谱曲线的红边产生了较大的影响，其他波段区内，相关系数均较低。上述结果表明成像光谱能够反映小麦幼苗干旱胁迫下的长势状况。4.利用高光谱成像监测小麦苗期的田间长势状况。获取大田正常生长的小麦的成像光谱，基于遗传算法挑选的6个特征波长，分别与小麦幼苗单株叶片的含氮量、叶绿素两个营养组分信息进行预测，采用PIS方法建立相关模型，进行相关分析结果表明：基于遗传算法结合PIS方法建立的小麦氮素、叶绿素GA-PLS模型，均优于基于全谱波长范围的FS-PLS模型和基于经验选择波长范围的ES-PLS模型。预测小麦苗期氮素的GA-PLS模型相关系数R=0.9，校正标准差SEC=0.0841，预测标准差RPD=1.4358，相对百分值RSEP=9.3893%。小麦叶绿素的GA-PLS模型相关系数R=0.8303，校正标准差SEC=0.1169，预测标准差RPD=0.8064，相对百分值RSEP=13.899%。5.利用高光谱成像识别小麦穗发芽程度。应用成像光谱仪分别对四种不同水分处理下的120个小麦穗进行分析，采集小麦穗部及籽粒的平均光谱反射率，并进行对比分析研究，结果表明：基于高光谱成像特征波段提取的RGB合成彩色效果图要优于基于机器视觉的小麦穗发芽成像图。小麦穗部发芽与否会在其光谱特性上有所反映，特别是在675nm处，发芽小麦与未发芽小麦的光谱反射率存在明显差异，小麦穗发芽的部位在675nm处会出现吸收谷，而未发芽部分则没有，以此通过光谱曲线在675nm处的变化情况，可判断小麦穗部是否发芽。从而说明，成像光谱可以较好的区分不同水分处理下的小麦穗发芽情况，表明利用高光谱成像技术对小麦穗发芽进行无损监测具有可行性。更多还原

【Abstract】 Wheat was the world’s most widely distributed crops. It is also one of the largest food cropacreage and among the largest food production in China. It was an important commodity grainand strategic reserves of staple food varieties. Seedling growth and development affectes thewheat’s final yield, quality and safety. So seedling vigor agronomists, breeders, biochemists, areconcerning about the crop characteristics. Chlorophyll content as an indicator of the cropphotosynthesis ability and the growing status. Currently, the monitoring of crop growing statuswere mostly focused on the middle and final growth period, where spectroscopy analysis iswidely used to capture group parameters. Seedling monitoring is focused more on the analysisof individual parameter; however it was lack of effective detection method. Hyperspectralimaging, known also as spectroscopic imaging, is an emerging technique that integratesconventional imaging and spectroscopy to attain both spatial and spectral information from anobject. It could obtain the spectrum of each local area, and has great advantage on differentiatingindividual differences. It has been used in aviation and satellite remote sensing for years.Recently, with the reducing of cost, the hyperspectral imager for field purpose is possible. Forthis study, we monitor the wheat seedling growth by hyperspectral imaging technology and highresolution image integration. In this paper, we used the hyperspectral imagers to detect thechlorophyll content for salt stress of the wheat seedling, freezing injury of wheat seedlings underlow-temperature stress, drought injury of wheat seedlings under low-water stress, recognition ofWheat Pre-harvest Sprouting, and the chlorophyll content of wheat seedings.1. An auto-development pushbroom imaging spectrometer (PIS) with wavelength range of400-1000nm was applied to measure the detection of chlorophyll content for salt stress of thewheat seedling by hyperspectral imaging. It showed that according to images of spectralimaging for leaves of Chinese Spring (Salt-sensitive), Zhouyuan9369(common and high-yield)and Changwu134(salt-tolerant) wheat seedling under salt stress, growth of salt-sensitiveChinese Spring wheat seedling was inhibited and it was feasible to carry out qualitativeanalysis.We collected the imaging spectrum of126wheat samples in the wavelength range of400~1000nm, selected the average spectrum, exerted Correlation Analysis on the spectrum ofwheat seedlings with the SPAD value, It could be seen that the biggest absolute value of thecorrelation coefficient was at693nm, which was considered as the characteristics wavelength ofwheat seedlings. To establish the linear regression model using this wavelength, and substitutedthe reflectance data of each point into the model, then we got the SPAD value of each point, toform the relative content distribution map of chlorophyll, whereby to diagnose the distribution of seedlings component. Partial least square regression (PLSR) method was used to build thecalibration model. Results showed that the extracted hyperspectral spectra had high correlationwith chlorophyll content. The correlation coefficient of the calibration model is R=0.8138, thestandard error of prediction is SEP=4.75.2. Diagnosis of freezing stress in wheat seedlings using hyperspectral imaging technology.Imaging spectrometry and image integration were used to study the extent of freezing injury onwheat seedlings on three dates in2010. Thirty wheat cultivars with limited freezing resistancethat are usually grown in Southern China were grown in pots located outdoors in Beijing.Imaging spectra of potted wheat samples were acquired in the spectral range of450nm to900nm, and a polyphenol tester was used to determine the nitrogen balance index of wheat samples.Increased freezing injury was related to increased spectral reflectance in the450nm to650nmwave band and decreased spectral reflectance in the700nm to900nm wave band. Averagespectral reflectance of wheat seedling canopies was negatively correlated (-0.7) with nitrogenbalance index in the red edge area between650nm and700nm. Absolute values of correlationcoefficients under freezing stress at three measuring dates reached a maximum at680nm, andthis wavelength was used as the characteristic wavelength for freezing-injury diagnosis of thewheat seedlings. From spectral images at this characteristic wavelength, it was feasible tointuitively observe the area and extent of freezing injured wheat seedlings. Our results show thatit is feasible to monitor freezing stress of wheat seedlings by use of hyperspectral imaging whichcould accurately reflect freezing-injured parts of wheat seedlings.3. Diagnosis of drought stress in wheat seedlings using hyperspectral imaging technology.The severity of the drought stress conditions different period of three acquisition imagingspectrometer data2011-11-15,2011-11-23,2011-12-01wheat seedlings, respectively, andrecord the corresponding seedling per plant leaf chlorophyll SPAD values, of wheat seedling as2011-11-23periods of drought stress aggravated Yangmai15Yangmai158, the the SPADaverage of Jingdong8monoclonal measured three poor drought resistance of potted wheatleaves begin to reduce small, and better drought resistance Nongda211measured SPAD valueof leaves of wheat seedlings monoclonal close to the measured values in the two periods isinsignificant. Analysis found that hyperspectral imaging technology on Wheat canopy spectracollected under different drought stress period in the wavelength range of400~1000nm,450nm~900nm spectral wavelength range2012-12-01wheat seedlings during drought stress isthe most serious wheat seedling, mean canopy spectral reflectance, and do not have thecharacteristic peaks of the spectral reflectance of green vegetation. Subsequently, seedlingdrought stress under the canopy spectral reflectance, first derivative, and the chlorophyll SPAD average of measured correlation coefficient curve, correlation analysis, we found that thespectral curve in the range of700~780nm red band within a positive effect related to asignificant, suggesting that drought stress had a greater impact may be the curve red edge, theother band area, significant correlation coefficients are very low. We can see from the abovefindings: Imaging Spectrometer PIS has the advantages of imaging and it maps thecharacteristics of unity, to extract the spectral characteristics, and to reflect the growingconditions in the wheat seedling drought stress.4. The application of hyperspectral imaging acquisition field of wheat seedling canopy andindividual plant leaf spectral imaging, based on genetic algorithm selected six characteristicbands corresponding to the characteristic wavelength, the leaves of wheat seedlings per plantwere measured nitrogennutritional component information the amount of chlorophyll tworepresentative wheat growing good or bad, true value, to build mathematical models usingpartial least squares (PIS), correlation analysis, the wheat seedlings per plant leaf componentinformationchoice of different spectral wavelength range of model results, the results show: thecombination of partial least squares (PIS) and the blades of wheat seedlings monoclonalnitrogen, chlorophyll value of the model based on genetic algorithm, are superior based on thefull spectrum of wavelengthsthe range of FS-PLS and ES-PLS model based on experience toselect the wavelength range. Comprehensive evaluation to predict wheat seedling nitrogencomponent information modeling results the correlation coefficient R=0.9, correction standarddeviation of the SEC=0.0841, standard error of prediction of the RPD=1.4358, the relativepercentile RSEP=9.3893%. At the same time, the leaf chlorophyll value of the wheat seedlingper plant based on genetic algorithm combined with partial least squares method to establish themodel results of the correlation coefficient R=0.8303, corrected standard deviation of the SEC=0.1169, standard error of prediction the RPD=0.8064, and the relative percentage value RSEP=13.899%.5. Study on recognition of wheat pre-harvest sprouting based on hyperspectral imagingtechnology. Imaging spectrometer was used to analyze120strains of wheatears under fourdifferent watering treatment ways. Comparative analysis was carried out for mean spectralreflectivity of ears and seeds of the four groups of wheat wheatear, and results showed thatextracted RGB color effect image based on characteristic wavelength of hyperspectral imagingwas better than imaging image of wheat pre-harvest sprouting based on machine vision.Whether wheat pre-harvest sprouting occurred could be reflected by spectral characteristics.Especially at675nm, there was a significant difference for spectral reflectivity betweensprouting wheat and non-sprouting wheat. Sprouting parts of wheatears presented absorption valley at675nm, while non-sprouting parts didn’t present it. Therefore, it was feasible to judgewheatears’ sprouting situations according to changes of spectral curve at675nm. Therefore, it isillustrated that imaging spectra can better differentiate wheat pre-harvest sprouting situationsunder different watering treatment ways, suggesting that it is completely feasible to usehyperspectral imaging technology to carry out nondestructive monitoring for wheat pre-harvestsproutingThe results above of the all, indicated that hyperspectral imaging were suitable for thenon-invasive detection of chlorophyll content of wheat seedling.更多还原