基于近红外技术和偏最小二乘判别法对卫青萝卜内部品质的快速判别

为实现天津卫青萝卜内部品质的快速无损检测,以天津卫青萝卜为试验对象,对其进行近红外光谱采集,并对其可溶性固形物、水分、糠心3个品质指标进行检测,根据光谱数据、品质指标构建偏最小二乘判别分析(PLS-DA)模型进行聚类判别。结果表明,经MA平滑处理后的样本光谱,在波长700～750 nm,900～950 nm范围内,样本的吸光度值范围较宽,因此选择在波长700～950 nm范围的光谱数据进行建模。卫青萝卜水分含量、可溶性固形物含量的平均值±样本标准差(x±s)分别为92.02%±0.01%,7.61±0.69°Brix。偏最小二乘法建立的模型具有良好的预测性,卫青萝卜可溶性固形物预测值和真实值的决定系数R2为0.823 6,水分含量R2为0.874 0,糠心模型准确率高达94.44%。由此可见,近红外光谱技术对天津卫青萝卜内部品质的快速无损检测具有可行性。     

寒富苹果可溶性固形物可见/近红外光谱无损检测模型的优化

为提高可见/近红外光谱无损检测寒富苹果可溶性固形物的检测精度,应用近红外漫反射光谱仪对寒富苹果进行扫描,对主成分回归(principal component regression,PCR)、偏最小二乘法(partial least squares,PLS)和改进偏最小二乘法(modified partial least squares,MPLS)三种建模方法进行比较,通过改变波长的范围、导数处理、去散射处理、标准化处理、加权多元离散校正及间隔平滑处理等光谱预处理,研究不同建模和光谱预处理方法对寒富苹果可溶性固形物可见/近红外光谱无损检测模型准确性的影响。结果表明,在780~1 100 nm范围内,采用MPLS,间隔点为2,平滑点为2,结合去散射处理和一阶求导处理所建立的寒富苹果可溶性固形物定标模型最好,其定标模型的校正交互验证标准误差(standard error of cross validation,SECV)为0.306,交互验证决定系数(determination coefficient of cross validation,R2cv)为0.961;预测标准偏差(square error of prediction,SEP)、预测决定系数(determination coefficient of prediction,R2P)、预测相对分析误差(residual predictive deviation,RPD)分别为0.357、0.944、4.967,表明模型具有良好的预测效果,适用范围广。建模方法、波长范围、导数处理、间隔平滑处理、去散射处理使模型误差分别降低了14.688%~53.407%、20.787%~33.146%、1.918%~13.123%、1.813%~7.553%、0~2.647%,建模方法和光谱预处理对模型优化的次序依次为:建模方法波长范围导数处理间隔平滑处理去散射处理。     

近红外光谱无损检测石榴中花色苷含量研究

以全国不同地区的97个石榴为样本,研究近红外光谱无损检测石榴中花色苷的含量,探讨了不同数据处理和回归方法对建模效果的影响。结果表明,对原始光谱进行一阶微分、标准多元离散校正法处理后,采用偏最小二乘法建立的石榴花色苷含量预测模型,预测偏差为0.148,预测标准差(SEP)为1.47,相关系数为0.829,模型预测良好,说明近红外光谱无损检测石榴的品质是可行的。     

利用VIS/NIR反射光谱快速检测柿饼加工中水分含量

水分是柿饼的重要组成成分,也是影响柿饼制作过程的重要因素。利用可见/近红外反射光谱对柿饼制作过程中的水分含量进行检测。首先,获取柿饼在不同加工阶段的可见/近红外反射光谱(400～1 000 nm),采用烘干法测定柿饼水分含量。然后,对光谱进行Mean smoothing (MS)平滑、多元散射校正(MSC)和一阶导数(1-D)预处理。最后,对不同预处理光谱,结合样本水分含量,使用Samples set partitioning based on joint x-y distance (SPXY)方法划分校正集和验证集,基于SPA方法选择特征波长,建立多元线性回归(MLR)预测模型。结果表明,反射光谱经过MS处理后,确定的9个最优波长组合建立水分检测模型的预测结果最好:预测相关系数(Rp)为0.969 0,预测标准残差(SEP)为3.472 9%,可见/近红外反射光谱技术可以较好地预测柿饼制作过程中的的水分含量。研究可为柿饼加工过程中的品质快速检测提供一定的技术支撑。     

基于近红外法的鲜食大豆品质快速分析技术

为建立一种鲜食大豆品质的非破坏性快速检测方法,探究近红外分析技术在鲜食大豆品质检测中应用的可行性,以四川省成都市郫都区种植的66份鲜食大豆样品为材料,采用FOSS近红外谷物分析仪扫描得到光谱,并对所有材料的水分、蛋白质、可溶性糖和粗脂肪含量进行常规实验室分析,利用偏最小二乘法(PLS)建立近红外光谱与化学实验数据的相关模型,并进行模型优化、验证。得到的模型预测范围分别为水分含量60.03%~71.28%、可溶性糖含量2.36%~7.81%、可溶性蛋白含量1.03%~8.56%、粗脂肪含量4.33%~7.60%,预测系数(0.95)较高,标准误差(1.0)较低,实验结果显示,利用近红外分析技术建立定标模型用于检测鲜食大豆品质是可行且可靠的,该方法可快速检测鲜食大豆品质且不损坏籽粒,可用于鲜食大豆的种质资源评价、品质分级等。     

基于 SPA-RBF神经网络的小麦蛋白质含量无损检测

传统半微量凯氏法测量小麦蛋白质含量繁琐费时,应用近红外光谱分析技术结合SPA-RBF神经网络对小麦蛋白质含量进行快速、无损检测.采用SPXY算法划分校正集和预测集样本,运用连续投影算法(SPA)对一阶微分和SNV预处理后的光谱数据提取敏感波点作为RBF神经网络的输入,建立小麦蛋白质含量的SPA-RBF神经网络校正模型.模型的预测均方根误差和预测相关系数可达到0.26576和0.975,预测效果较好,基本上可以完成粮食储备和食品加工行业对小麦及其制品品质的划分以及育种上的前期世代筛选.研究表明:近红外光谱技术结合SPA-RBF神经网络可实现对小麦蛋白质含量的检测,满足现代农业发展对小麦无损、实时、大量检测的需要.     

基于化学计量学的近红外光谱法检测白酒中酯含量研究

为实现快速检测白酒中4种主要酯类物质的含量,采用化学计量学方法对模拟白酒样品的近红外光谱数据进行模型构建,并用白酒样品进行验证试验。结果表明,模型预测值与实际值的相关系数均大于0.97。使用白酒样品进行验证试验,预测值与测定值无显著差异。说明采用近红外光谱进行白酒中酯类物质检测可以实现快速、无损、多参数、多指标检测。     

香蕉成熟期品质可视化与高光谱成像研究

为解决香蕉采后转色快和易腐烂而导致的果品等级下降问题,利用高光谱成像技术建立快速无损检验香蕉果实成熟度的预测方法.测定了香蕉在20、25、30℃三种不同贮藏温度下的可溶性固形物含量、含水率和硬度的动态变化,并与同期对应的光谱数据进行对比分析.同时,利用蒙特卡洛偏最小二乘法(MCPLS)选取其特征波长,并建立多元线性回归(MLR)预测模型.结果表明:香蕉果实的可溶性固形物含量、含水率和硬度的预测模型的决定系数分别为0.8584、0.8735和0.9128,实现了不同成熟阶段香蕉果实的可溶性固形物含量、含水率和硬度的无损评价.将高光谱成像技术应用于香蕉果实成熟期品质参数的快速无损检验具有良好可行性.     

向日葵籽仁脂肪和脂肪酸含量近红外光谱模型的建立

为实现向日葵品质的快速无损检测,选取50份具有代表性的油用向日葵材料,采用偏最小二乘法（PLS）构建籽仁脂肪、亚油酸、油酸、硬脂酸和棕榈酸含量的近红外光谱（NIRS）模型。结果表明,脂肪、亚油酸、油酸含量模型校正和验证相关系数均大于0.96,且预测值与化学值相对误差均在10%以下,能够达到样品成分含量的快速测定。硬脂酸和棕榈酸含量模型校正相关系数分别为0.92和0.82,验证相关系数分别为0.83和0.74,预测值与化学值相对误差在4.66%~17.99%之间,可用于样品成分含量的初步预测。本研究构建的NIRS模型,有助于油用向日葵种质资源品质鉴定和快速筛选。     

可见-近红外光谱对花椰菜色泽的检测研究

花椰菜色泽变化和品质密切相关,利用可见-近红外光谱技术对花椰菜的色泽L*值进行研究。结果表明,归一化为最佳预处理方法,使用连续投影算法提取光谱数据特征波长并建立偏最小二乘回归模型。预测模型的相关系数达到了0.908 9,预测均方根误差为0.541 1。说明可见-近红外光谱技术检测花椰菜的色泽L*值变化是可行的。     

近红外光谱法测定大米中的淀粉含量

用化学方法测定64个大米样品中的淀粉含量,利用近红外谷物分析仪采集样品的近红外光谱,选择合适的光谱区间和光谱预处理方法。50个定标集样品的近红外光谱经二阶导数及标准多元离散校正(Standard MSC)预处理,结合偏最小二乘法(PLS)建立了大米中的淀粉含量测定的定标模型,其相关系数为0.8780。14个验证集样品用于外部检验,大米中的淀粉含量的模型预测值与化学值之间的相关系数为0.9498。     

超声法同时测定牛奶中脂肪和蛋白质含量

采用超声法测量牛奶成分,并对牛奶超声特性进行了探讨,分别通过偏最小二乘法、非线性偏最小二乘法和非线性主成分回归,建立脂肪和蛋白质的校正模型。结果表明,用非线性偏最小二乘法建模预测的结果最好,脂肪的预测标准偏差RMSEP值为0.2533,蛋白质的预测标准偏差RMSEP值为0.2303,而与标准化学法相比,其预测结果不存在明显差异。     

近红外光谱法快速测定白酒中的酒精度

为了得到白酒工业中酒精度的快速检测技术,将偏最小二乘法与傅立叶变换近红外光谱法相结合,建立白酒酒精度的快速定量模型。通过标准归一化预处理光谱,光谱范围选择5731.40~5897.25、5901.11~6063.10、8327.12~8423.54 cm-1,主成分数为5,得到模型的内部交互验证相关系数(R)为0.9992,交互验证均方差(RMSECV)为0.263;模型的预测值与实测值的相关系数为0.99,预测标准偏差(RMSEP)为0.435。结果表明,模型的预测效果很好,具有较高的精密度和良好的稳定性,能满足生产中白酒酒精度的快速检测要求。     

近红外快速测定干辣椒中辣度的研究

以干辣椒为对象,采用近红外快速测定方法检测其辣度。首先,使用高效液相色谱法对8种干辣椒中辣椒碱类物质含量进行准确测定,确定了定量指标辣度。然后,采集干辣椒粉样品的近红外光谱数据,利用偏最小二乘法(PLS)建立检测模型,并对检测波长范围及模型主因子数进行了筛选。结果表明,使用PLS进行模型的建立,校正集方程相关系数0.987 1,验证集方程相关系数0.870 4;校正均方根误差2 870,交叉验证均方根误差9 476,主因子数为8。最终得到的检测模型能够满足对干辣椒中辣度的快速检测要求,且具有较好的准确度。     

Investigating the effects of simulated transport vibration on tomato tissue damage based on vis/NIR spectroscopy

Visible and near infrared (vis/NIR) spectroscopy combined with chemometrics were investigated to evaluate the effects of simulated transport vibration levels on damage of tomato fruit. A total of 280 tomato samples were randomly divided into 5 groups; each group was subjected to vibration at different acceleration levels. A total of 230 samples (46 from each group) were selected as a calibration set; whereas 50 samples (10 from each group) were selected as a prediction set. Raw spectra, differentiation (the first derivative) spectra, extended multiplicative scatter correction (EMSC) processed spectra and standard normal variant combined with detrending (SNV–DT) processed spectra were used for calibration models. SNV–DT processed spectra had the best performance using for partial least squares (PLS) analysis. The PLS analysis was implemented to calibrate models with different wavelength bands including visible, short-wave near infrared (SWNIR) and long-wave near infrared (LWNIR) regions. The best PLS model was obtained in the vis/NIR (600–1600 nm) region. Using a grid search technique and radial basis function (RBF) kernel, four least squares support vector machine (LS–SVM) models with different latent variables (7, 8, 9, and 10 LVs) were compared. The optimal model was obtained with 9 LVs and the correlation coefficient (r), root mean square error of prediction (RMSEP) and bias for the best prediction by LS–SVM were 0.984, 0.137 and 0.003, respectively. The results showed that vis/NIR spectroscopy could be applied as a reliable and rapid method for predicting the effect of vibration levels on tissue damage of tomato fruit.     

应用近红外光谱技术分析稻米蛋白质含量

以稻谷、米粒、米粉3种形态的样品，应用近红外光谱技术(NIRS)和偏最小二阶乘法(PLS)，建立了6个稻米蛋白质含量近红外光谱数学模型，并对模型预测结果的准确性进行了评价。结果表明，糙米蛋白质含量的稻谷、糙米粒和糙米粉近红外光谱预测模型校正决定系数(R_C²)分别为0.893、0.971和0.987，校正标准差(RMSEC)分别为0.507、0.259和0.183；精米蛋白质含量的稻谷、精米粒和精米粉近红外光谱预测模型R_C²分别为0.897、0.984和0.986，RMSEC分别为0.497、0.186和0.190。模型内部交叉验证分析表明，预测糙米蛋白含量的稻谷、糙米粒和糙米粉模型内部交叉验证决定系数(R_CV²)分别为0.865、0.962和0.984，内部验证标准差(RMSECV)分别为0.557、0.290和0.205；预测精米蛋白含量的稻谷、精米粒和精米粉的模型R_CV²分别为0.845、0.951和0.979，RMSECV分别为0.594、0.316和0.233。模型外部验证分析表明，预测糙米蛋白含量的稻谷、糙米粒和糙米粉近红外光谱模型外部验证决定系数(R_V²)分别为0.683、0.801和0.939，外部验证标准差(RMSEV)为0.962、0.799和0.434；预测精米蛋白含量的稻谷、精米粒和精米粉近红外光谱的模型R_V²分别为0.673、0.921和0.959，RMSEV为0.976、0.513和0.344。用米粉建立的近红外光谱预模型准确性最高，米粒次之，基于稻谷的预测模型准确性相对较低；内部交叉验证和外部验证表明，近红外光谱分析技术与化学分析方法一致性较好，且能保证样品的完整性，在水稻优质育种和稻米品质分析中具有广泛的应用价值。     

Reflectance estimation of canopy nitrogen content in winter wheat using optimised hyperspectral spectral indices and partial least squares regression

Many spectral indices have been proposed to derive plant nitrogen (N) nutrient indicators based on different algorithms. However, the relationships between selected spectral indices and the canopy N content of crops are often inconsistent. The goals of this study were to test the performance of spectral indices and partial least square regression (PLSR) and to compare their use for predicting canopy N content of winter wheat. The study was conducted in cool and wet southeastern Germany and the hot and dry North China Plain for three winter wheat growing seasons. The canopy N content of winter wheat varied from 0.54% to 5.55% in German cultivars and from 0.57% to 4.84% in Chinese cultivars across growth stages and years. The best performing spectral indices and their band combinations varied across growth stages, cultivars, sites and years. Compared with the best performing spectral indices, the average value of the R² for the PLSR models increased by 76.8% and 75.5% in the calibration and validation datasets, respectively. The results indicate that PLSR is a potentially useful approach to derive canopy N content of winter wheat across growth stages, cultivars, sites and years under field conditions when a broad set of canopy reflectance data are included in the calibration models. PLSR will be useful for real-time estimation of N status of winter wheat in the fields and for guiding farmers in the accurate application of their N fertilisation strategies.     

A Dynamic Model for Water and Nitrogen Limited Growth in Spring Wheat to Predict Yield and Quality

Annual variation of bread wheat ( Triticum aestivum L.) yield and quality has caused problems for agronomic policy in northern regions. Yield prediction methods based on visual assessment of crop may be inaccurate as they are not based on quantitative data. The aim of this study was to develop a simple dynamic model, based on daily climatological data, enabling prediction of crop growth, and changes in crop yield, and grain protein concentration and starch quality. The model was built using field data collected in 1972–88. Spring wheat cultivars included in the study were Kadett and Ruso. The calibration of growth and Hagberg falling number (used as a measure of starch quality) sub-models resulted in a highly significant positive correlation between measured and calculated values. The calibration of nitrogen sub-models failed, however, with poor correlation between measured and calculated values. The model was tested against independent field data collected during 1989–90, and results correlated with calibration results. The yield predictions based on independent field data were accurate, and the same as or similar to field trial results. However, the independent Hata revealed flaws in soil-water and Hagberg falling number sub-models.     

The effect of the chemical composition of a ferrallitic soil on neutron probe calibration

In a ferrallitic soil in French Guiana, the neutron probe calibration appeared to be problematic: considerable variations in the neutron count rate were observed at very short range and with an almost constant volumetric water content. This local variability of the count rate was explained by the mineralogical heterogeneity of the schist weathering horizon where subverticaly oriented layers are especially rich in boron, element with large thermal neutron absorption cross section. Various calibration methods were carried out and their limits were pointed out. The field gravimetric calibration without taking into account the soil physical and chemical spatial variations appeared to be risky, even if different pedological horizons are considered separately. A calibration based on the neutron absorption Σa and diffusion Σd cross sections calculated from chemical analysis led to overestimates of the volumetric water content. This could be explained by the concentration of boron atoms in sand-size tourmaline crystals which reduces their neutron absorption properties. The direct measurement of thermal neutron absorption and diffusion cross sections on soil samples in a graphite pile seems to be the best calibration procedure, but it has to be repeated as often as the spatial variability required.     

利用混合蛙跳算法优化基于APSIM的旱地小麦产量形成模型参数

模型参数的快速、准确估算是产量形成模型应用的重要前提。在基于APSIM (agricultural production systems simulator)的旱地小麦产量形成模型参数本土化率定过程中, 存在体量大、耗时长、精度低、效率低的缺点, 本研究利用智能算法优化模型参数, 试图解决上述问题。依据甘肃省定西市安定区李家堡镇麻子川村2002—2005年、凤翔镇安家沟村2015—2016年大田试验数据以及定西市安定区1971—2016年气象和产量资料, 运用混合蛙跳算法分组轮换和全局信息交换的智能策略, 对基于APSIM的旱地小麦产量形成模型参数进行了优化, 并采用相关性分析方法检验。该优化方法利用青蛙智能的群体生物进化学习策略, 可实现对小麦产量形成模型参数的估算, 较APSIM平台参数本土化率定常用的穷举试错法, 参数优化后产量模拟精度显著提高, 均方根误差(RMSE)平均值由79.13 kg hm ^-2降低到35.36 kg hm ^-2, 归一化均方根误差(NRMSE)平均值由5.97%降低到2.63%, 模型有效性指数(M_E)平均值由0.939提高到0.989。该方法全局优化能力强, 收敛速度快。