近红外光谱变量筛选提高西瓜糖度预测模型精度

水果的内部品质是水果分级、保鲜及存储的一项重要指标,利用近红外光谱技术对西瓜内部品质进行快速无损检测研究有着非常重要的意义。为了研究变量筛选方法对西瓜糖度预测模型精度的影响,该文以麒麟瓜为研究对象,利用近红外漫透射光谱技术对麒麟瓜可溶性固形物含量（SSC）进行检测,采用偏最小二乘回归（PLSR）,多元线性回归（MLR）和主成分回归（PCR）建立麒麟瓜可溶性固形物数学模型,并探讨等间隔平均光谱和等间隔抽取光谱变量筛选结合连续投影算法（SPA）对预测模型精度的影响。研究结果表明：光谱经等间隔抽取（间隔5,115个变量）经归一化预处理,结合SPA优选出6个波长建立的PLSR预测模型的相关系数（rpre）为0.828、校正均方根误差（RMSEC）为0.589、预测均方根误差（RMSEP）为0.611。该模型预测效果相对较优,建模时间短,提高了模型的预测能力和预测精度。该研究为西瓜内部品质的在线无损检测提供研究基础。     

玉米种子活力近红外光谱智能检测方法研究

为了实现玉米种子活力的快速无损检测,提出利用近红外光谱和BP神经网络来建立玉米种子活力智能检测模型。首先通过人工老化将样本按老化程度分为3种级别,采集样本的近红外光谱。分别通过卷积平滑(S-G)和多元散射校正(MSC)及二者组合的方法消除光谱噪声和去除奇异光谱。然后分别用主成分分析(PCA)和离散多带小波变换(DWT)提取光谱特征,作为BP神经网络的输入。依据预处理及特征提取的不同构建出6种BP神经网络种子活力检测模型。试验结果表明,组合预处理方法与主成分分析特征提取结合构建的模型最优,其识别的准确率为95.0%,平均识别时间为26.25ms。研究结果为玉米种子活力的快速无损检测提供了理论依据和实用方法。     

基于近红外光谱技术的大米品种快速鉴别方法

为探索大米无损检测技术,提出了一种基于可见-近红外光谱技术快速、无损鉴别大米品种的新方法。首先采用主成分分析法对大米品种进行聚类,然后利用小波变换技术提取光谱特征信息,把光谱特征信息作为人工神经网络的输入建立品种识别模型,对大米品种进行鉴别。从每种大米60个样本共计180个样本中随机抽取150个样本（每种50个样本）用来建立神经网络模型,剩下的30个大米样本用于预测。品种识别准确率达到100%。说明所提出的方法具有很好的分类和鉴别作用,为大米的品种鉴别提供了一种新方法。     

基于漫反射光谱的茶园土壤硝态氮检测

该文研究了充分利用土壤漫反射光谱在可见-近红外波段的有效信息,研究快速准确检测土壤硝态氮含量的新方法。试验选取89个风干土壤样本,经粉碎过直径1 mm筛孔后,使用 FieldSpec 3便携式光谱仪（光谱波长范围：400～2 500 nm）,获取其漫反射光谱。检查各土样的原始光谱的有效性并进行平均,经偏最小二乘法partial least squares（PLS）聚类分析后,选取其中的63个样本构成校正集建立模型,10个样本构成预测集进行模型验证。通过一阶微分与滑动平均滤波相结合的预处理方法,用15个主成分建立的主成分+神经网络模型为最好,其校正模型的回判相关系数为0.9908,均方根误差（RMSEC）为1.4528,预测模型的相关系数为0.7179。研究结果表明,利用可见-近红外光谱技术可以准确地检测茶园土壤硝态氮含量。     

柑桔黄龙病近红外光谱无损检测

为探讨快速无损检测柑桔黄龙病的可行性,应用近红外光谱技术结合机器学习方法进行研究。在4000~9000cm-1光谱范围内,采集黄龙病、缺素和健康3类叶片样本的近红外光谱。采用一阶导数、平滑和多元散色校正组合的光谱预处理方法,消除光谱的基线漂移和散射效应。分别对偏最小二乘判别模型(PLS-DA)的主成分因子数和最小二乘支持向量机(LS-SVM)的输入变量数量、核函数类型及其参数进行了优化,建立了PLS-DA和LS-SVM模型。采用预测集样本,评价模型的预测能力,经比较,采用11个主成分得分向量为输入、线性核函数和惩罚因子为2.25的LS-SVM模型预测效果最佳,模型误判率为0。结果表明采用近红外光谱技术结合最小二乘支持向量机进行柑桔黄龙病无损检测是可行的。     

基于电学参数的苹果可溶性固形物含量预测

为实现苹果可溶性固形物含量的无损检测，该研究提出了一种长短期记忆编解码和多层感知机（LSTMED-MLP,long short-term memory encoder-decoder-multi-layer perceptron）融合的介电特征预测方法。在0.158～3 980 kHz频率范围内的9个频率点下，采用介电谱测量仪获取300个富士苹果的电学参数，其中每个频率点对应15项电学参数，即每个苹果对应135项电学特性参数，之后通过苹果基因组学理化分析方法，获取可溶性固形物含量；根据电学参数与可溶性固形物含量，构建苹果关键基因组学参数的回归预测模型。为简化模型输入，提取样本变量特征，使用主成分分析（principal component analysis,PCA）和LSTMED模型，提取每个样本的40项特征值，作为非线性回归模型多层感知机（MLP）和XGBoost的输入，建立可溶性固形物含量预测模型。试验结果表明，LSTMED具有更好的适用性，且LSTMED-MLP模型的预测效果最好，在校正集和预测集上，决定系数分别为0.95和0.90，均方根误差分别为0.77和0.84，且对不同种...     

识别不同水稻株型的高光谱模式方法的建立

提出了一种用高光谱技术快速识别不同水稻株型的新方法。首先在试验田内选择33个不同的水稻品种,测定了每个品种的14个株型特征参数,并采用荷兰Avantes公司的AvaSpec-2048便携式光谱仪采集不同株型水稻的高光谱数据。通过聚类分析,将所有水稻品种分为差异较大的3个株型类别。再采用平均平滑法和标准归一化方法对光谱数据进行预处理,对光谱数据主成分分析并获得各主成分数据。最后将主成分数据作为BP神经网络的输入变量,株型类别作为输出变量,建立了三层人工神经网络识别模型,并用模型对预测样本进行预测。结果表明,预测准确率为100%。该方法实现了对不同水稻株型的快速、无损识别。     

基于极限学习机利用可见-近红外光谱数据判别土壤类型的方法研究

为了实现土壤类型的快速无损识别,提出了一种利用可见-近红外光谱、基于极限学习机的土壤类型鉴别方法。首先,获取4种不同类型土壤的320个样本波长在325～1075 nm范围内的可见-近红外光谱数据;其次,用主成分分析的数学方法对数据进行降维处理,最终提取了三个主成分来代表原光谱数据;再次,将320个样本的数据随机分为测试集和预测集两个部分,建立极限学习机模型,利用该模型对土壤类型进行识别。实验结果表明,将极限学习机应用于土壤类型的识别精度可达100%,其训练速度和泛化性优于BP神经网络和支持向量机,能够快速、准确、无损鉴别土壤类型,使用方便,具有推广价值。     

可见/近红外光谱快速鉴别米粉辐照剂量

为了实现对不同剂量辐照处理后米粉的快速鉴别,提出了一种基于可见-近红外光谱技术的快速、无损检测方法。试验先利用不同剂量的60Coγ-射线对米粉进行辐照处理,得到了4种样品共200个样本。再应用ASD可见-近红外光谱仪获取所有样本的反射光谱数据,并采用主成分分析方法对数据进行聚类分析,将提取的前6个主成分作为BP神经网络的输入值,建立不同米粉样品的鉴别模型。结果表明,在设定偏差标准为±0.1的情况下,利用该模型对预测集样本进行鉴别,识别率达到100%。该文提出的方法具有很好的分类和鉴别作用,为快速鉴别米粉类产品是否经辐照灭菌及处理剂量等提供了新的技术方法。     

霉变稻谷脂肪酸含量的光谱检测模型构建与优化分析

为了实现霉变稻谷脂肪酸含量无损、快速检测,该文研究应用可见/近红外光谱技术检测霉变稻谷的脂肪酸含量。考虑到直接选用霉变稻谷可见/近红外光谱数据构建脂肪酸含量预测模型存在建模费时、预测失准等问题,研究提出了霉变稻谷脂肪酸含量的可见/近红外优化校正模型。研究中通过光谱-理化值共生距离(sample set partitioning based on joint xy distance,SPXY)算法结合偏最小二乘法初步分析了不同校正集样本预测霉变稻谷脂肪酸含量的差异;利用连续投影算法(SPA)提取了反映霉变稻谷脂肪酸含量变化的特征波段;采用偏最小二乘法(partial least square,PLS)和多元线性回归法(multivariable linear regression,MLR)分别建立了基于特征波段光谱反射值的霉变稻谷脂肪酸含量预测模型,并对比分析了采用SPXY样本集划分的模型预测效果。结果表明:采用SPXY法筛选出的65个校正集样本分布与初始校正集相近,脂肪酸含量变化范围为18.55~127.26 mg,其标准差为32.39;SPA算法最终从256个全光谱波段中优选出9个特征波段,实现了光谱数据的压缩;分别建立的SPXY-SPA-PLSR模型和SPXY-SPA-MLR模型预测霉变稻谷脂肪酸含量相关系数RP为0.922 1和0.915 9,预测均方根误差RMSEP为13.889 3和14.261 0;SPXY筛选校正集所构建模型预测精度与初始校正集所建模型相当,但校正集样本数量减少为初始校正集的41%,运算时长缩短为初始样本集的32%,提高了模型的校正速度。     

黄土的红外反射光谱与红外光声光谱特征及其差异研究

红外光谱法作为一种新的研究手段已经广泛应用于土壤分析,由其检测区域和手段的不同又可分为多种光谱类 型。本研究以第四纪黄土为例,系统地比较了近红外区和中红外区反射光谱和光声光谱的吸收特征及其差异。结果表明,中红外光谱比近红外光谱的信息更为丰富,且中红外光谱与样品中物质的特征吸收关系更加密切,从而更有利于土壤定性与定量分析。土壤的反射光谱和光声光谱表现出了明显不同的特征,在近红外区,反射光谱和光声光谱吸收明显不同,而在中红外区,反射光谱和光声光谱具有相对应的吸收,但相对吸收强度明显不同,且吸收峰的位置也发生改变,尤其在1 000 ~ 2 000 cm-1谱区,反射光谱相互干扰很强,而光声光谱的吸收特征更为明显。在黄土的分类鉴别上,反射光谱优于光声光谱。红外反射光谱和光声光谱在不同波段下具有不同的吸收灵敏度,在土壤定性与定量分析中各自都将具有其明显的优势。     

拉曼光谱在精细农业土壤成分快速检测中的研究进展

拉曼光谱分析技术利用分子运动对入射光产生非弹性散射的原理对分子成分进行检测,具有受水分干扰小、样本预处理小、与红外光谱信息互补等特点,在土壤成分快速分析方面展现了很大的优势。但是拉曼光谱信号弱,受荧光干扰强,为土壤拉曼信号的有效获取带来困难。为了分析拉曼光谱在土壤成分检测中的应用潜力,该研究综述了移频激发差分拉曼光谱技术、共焦显微拉曼技术以及表面增强技术等基于拉曼光谱的土壤成分检测技术,分析了土壤成分拉曼光谱检测的研究进展,并提出进一步研究建议。结果表明：1)脂肪族化合物以及芳香族化合物都具有拉曼活性,为基于拉曼光谱的土壤有机质含量的定性、定量分析提供了理论依据。为了弥补拉曼光谱对有机质整体定量预测精度的不足,采用红外-拉曼光谱融合方式补偿单独拉曼光谱数据中缺失的土壤有机质信息,可显著改善预测精度。2)利用表面增强技术可以增强土壤溶液中可溶性氮与土壤有效氮拉曼特征波峰的强度,获得了良好的定量预测效果,回归模型决定系数R²达到0.91～0.99。3)土壤中很多含磷的化合物都具有拉曼活性,拉曼光谱是识别土壤中不同磷酸盐形态的极其有效的工具,在土壤磷素含量的分析中,应用...     

红外光谱在土壤学中的应用

红外光谱技术在土壤学中已得到较广泛的应用,它能够综合地反映土壤体系的物质组成及其相互作用,为研究土壤中物质循环及其作用过程提供了新的手段。本文回顾了近年来红外光谱技术在土壤学中的应用,包括透射光谱在土壤定性分析中的应用,并重点介绍红外反射光谱与化学计量学相结合的光谱建模技术发展情况及其在土壤定量分析中的应用。同时本文探讨了基于光声效应的红外光声光谱技术,红外光声光谱非常适合用于土壤这种复杂、非透明体系的研究,能够克服传统透射和反射光谱中存在的缺陷,测定快速方便,并具有较高的灵敏度和测量精确度,具有很大的应用潜力。     

Compositional and functional features of humic acids from organic amendments and amended soils in Minnesota, USA

The use of organic amendments requires an adequate control of the chemical quality of their humic acid (HA)-like fractions and of the effects that these materials may have on the status, quality, chemistry and functions of native soil HAs. In this work, the compositional, functional and structural properties of the HA-like fractions isolated from a liquid swine manure (LSM), a municipal sewage sludge (SS), and two municipal solid waste composts (MSWCs) were evaluated in comparison to those of HAs isolated from three unamended soils and from the corresponding soils amended with LSW, SS, and MSWC at various rates in three field plot experiments conducted in Minnesota, USA. With respect to the unamended soil HAs, the HA-like fractions of the three amendments featured a greater aliphatic character, a marked presence of proteinaceous, S-containing and polysaccharides-like structures, an extended molecular heterogeneity, small organic free radical contents and a small degree of humification. The MSWC-HAs featured a larger degree of humification than LSM-HA and SS-HA. The three amendments affected in different ways and by various extents the compositional, structural and functional properties of soil HAs depending upon the nature, origin and application rate of the amendment. In general, the data obtained suggested that proteinaceous, S-containing and aliphatic structures contained in HA-like fractions of organic amendments were partially incorporated into native soil HAs.     

污泥堆肥和热干燥对土壤和土壤腐殖酸性质的影响

The effect of annual additions of composted sewage sludge (CS) and thermally dried sewage sludge (TS) at 80 t ha^-1 on soil chemical properties was investigated for three years in a field experiment under semiarid conditions. Humic acids (HAs) isolated by conventional procedures from CS, TS, and unamended (SO) and sludge amended soils were analysed for elemental (C, H, N, S and O) and acidic functional groups (carboxylic and phenolic) and by ultraviolet-visible, Fourier transform infrared and fluorescence spectroscopies. With respect to CS, TS had similar pH and total P and K contents, larger dry matter, total organic C, total N and C/N ratio and smaller ash content and electrical conductivity. Amendment with both CS and TS induced a number of modifications in soil properties, including an increase of pH, electrical conductivity, total organic C, total N, and available P. The CS-HA had greater O, total acidity, carboxyl, and phenolic OH group contents and smaller C and H contents than TS-HA. The CS-HA and TS-HA had larger N and S contents, smaller C, O and acidic functional group contents, and lower aromatic polycondensation and humification degrees than SO-HA. Amended soil-HAs showed C, H, N and S contents larger than SO-HA, suggesting that sludge HAs were partially incorporated into soil HAs. These effects were more evident with increasing number of sludge applications.     

Use of near‐ and mid‐infrared spectroscopy to distinguish carbon and nitrogen originating from char and forest‐floor material in soils

The presence of relatively inert organic materials such as char has to be considered in calibrations of soil C models or when calculating C‐turnover times in soils. Rapid and cheap spectroscopic techniques such as near‐infrared (NIRS) or mid‐infrared spectroscopy (MIRS) may be useful for the determination of the contents of char‐derived C in soils. To test the suitability of both spectroscopic techniques for this purpose, artificial mixtures of C‐free soil, char (lignite, anthracite, charcoal, or a mixture of the three coals) and forest‐floor Oa material were produced. The total C content of these mixtures (432 samples) ranged from 0.5% to 6% with a proportion of char‐derived C amounting to 0%, 20%, 40%, 50%, 60%, or 80%. All samples were scanned in the visible and near‐IR region (400–2500 nm). Cross‐validation equations for total C and N, C and N derived from char (C_char, N_char) and Oa material were developed using the whole spectrum (first and second derivative) and a modified partial least‐square regression method. Thirty‐six samples were additionally scanned in the middle‐IR and parts of the near‐IR region (7000–400 cm^–1 which is 1430–25,000 nm) in the diffuse‐reflectance mode. All properties investigated were successfully predicted by NIRS as reflected by RSC values (ratio of standard deviation of the laboratory results to standard error of cross‐validation) > 4.3 and modeling efficiencies (EF) ≥ 0.98. Near‐infrared spectroscopy was also able to differentiate between the different coals. This was probably due to structural differences as suggested by wavelength assignment. Mid‐IR spectroscopy in the diffuse‐reflectance mode was also capable to successfully predict the parameters investigated. The EF values were > 0.9 for all constituents. Our results indicated that both spectroscopic techniques applied, NIRS and MIRS, are able to predict C and N derived from different sources in soil, if closed populations are considered.     

应用拉曼和中红外衰减全反射光谱测定溶液和土壤中的硝酸盐

为比较拉曼光谱和红外光谱在溶液和土壤中硝酸盐含量定量分析的适用性,采用两种光谱对溶液和土壤中的NO3–-N含量(0~200 mg/L)进行快速测定。结果表明,溶液中硝酸盐的拉曼特征峰在1 047 cm–1处,该特征峰强度与NO3–-N浓度成正比,对1 035~1 060 cm-1波段拉曼光谱峰面积和NO3–-N含量进行线性回归,决定系数R2为0.995 4;溶液中硝酸盐的中红外衰减全反射光谱特征吸收峰在1 350 cm–1,吸收峰与NO3–-N含量成正比,特征吸收区1 200~1 500 cm–1峰面积与NO3–-N含量的决定系数R2为0.991 1,表明两种光谱都可用于溶液中硝酸盐的测定。对于土壤样品,红外光谱在1 250~1 500 cm–1处有硝酸盐吸收峰,且吸收峰与NO3–-N含量成正比,峰面积与NO3–-N含量之间的决定系数R2为0.968 4;而对于拉曼光谱,硝酸盐的拉曼峰因受较强干扰导致吸收峰不明显,峰面积与NO3–-N含量之间的决定系数R2仅为0.000 9,表明中红外衰减全反射光谱可用于土壤中硝酸盐的测定,而拉曼光谱则很困难。因此,拉曼光谱和中红外衰减全反射光谱都可用于溶液中硝酸盐的测定,且前者灵敏度要高于后者;中红外衰减全反射光谱可用于土壤中硝酸盐的测定,而拉曼光谱难以用于土壤中硝酸盐定量分析,这为硝酸盐的快速测定提供理论依据和技术支持。     

Properties of dissolved organic matter related to soil organic matter quality and nitrogen additions in Norway spruce forest floors

The quality of dissolved organic matter (DOM) is highly variable and little information is available on the relation of DOM quality to the structure and composition of its parent soil organic matter (SOM). The effect of increasing N inputs to forest soils on the structure and composition of both SOM and DOM also remains largely unclear. Here we studied the release of DOM, its specific UV absorption and two humification indices (HIX) derived from fluorescence spectra from Oa material of 15 North- and Central-European Norway spruce (Picea abies (L.) Karst.) stands. The Oa material was incubated aerobically at 15 °C and water holding capacity over a period of 10 months and extracted monthly with an artificial throughfall solution. Soil respiration was determined weekly. The influence of mineral N inputs on composition of DOM and on respiration rates was investigated on periodically NH₄NO₃-treated Oa samples of eight selected sites. Release of dissolved organic carbon (DOC) from untreated Oa material samples ranged from 0.0 to 58.6 μg C day⁻¹ g C⁻¹ and increased with increasing C-to-N ratio. One HIX and UV absorption of DOM were negatively correlated to the degree of oxidation of lignin-derived compounds and positively to the C-to-N ratio and – HIX only – to the aromatic C content of SOM. Mineral N addition had no distinct effect on respiration rates. In six of eight samples the N-treatment caused an increase in specific UV absorption or one HIX of DOM. However, these effects were not statistically significant. Addition of mineral N did not affect the rates of DOM release. Our results show that properties of SOM largely determine the amount and quality of DOM in forest floors. Changes of DOM quality due to mineral N additions are likely, but we cannot confirm significant changes of DOM release.     

基于光子传输模拟的苹果品质高光谱检测源探位置研究

光谱无损的检测方法是质检测最常用的方法之一。传统的光谱仪光源探头位置和源探距离相对固定,导致品质检测精度受限。为解决这个问题,提出基于蒙特卡洛的苹果多层组织的光子传输模拟,分析了光子入射最佳位置和源探距离,并用点光源高光谱仪实际拍摄红富士苹果进行验证。分析表明,光子在苹果赤道位置入射,具有73.12%概率到达更深的深度。源探距离与苹果的光学参数有关,形状为圆环,源探距离内外半径为1.5～10.15 mm。点光源高光谱仪采集红富士苹果的光谱信息,光子入射位置为赤道,源探距离为距离光源点半径2.7～11.7mm的圆环,与模拟数据分析结果基本一致。蒙特卡洛光子传输模拟方法为研究高光谱苹果品质无损检测开辟了新思路,分析结果可以为研究高光谱品质检测试验设计和苹果便携式品质检测光学仪器设计提供理论基础。     

Near-Infrared Reflectance Spectroscopy as a Method for Determining Organic Carbon Concentrations in Soil

Near-infrared reflectance spectroscopy (NIRS) has the potential to be a reliable method for accurately quantifying soil organic carbon (SOC). The objective of this study was to evaluate NIRS as a method for predicting SOC. Partial least squares (PLS) regression was used to predict SOC from soil reflectance values or the first derivative of the reflectance values. Two model validation techniques were evaluated: One was a full cross-validation and in the other 30 percent of the samples were removed from the calibration data set and then tested using the calibrated model. Significant relationships were observed for predicted SOC when compared to laboratory-measured SOC for all models evaluated, regardless of validation technique. The prediction models using the first derivative of the reflectance values outperformed prediction models using the reflectance values alone. In conclusion, NIRS can be used as a quick and accurate method for measuring SOC.