基于便携式近红外光谱仪的蓝莓热风干燥过程模型的建立

该文利用便携式近红外光谱技术采集蓝莓干燥过程光谱，以“蓝丰”蓝莓为研究对象，构建蓝莓热风干燥过程含水率快速检测模型。通过主成分分析法对蓝莓干燥阶段进行定性监测，分析比较竞争性自适应重加权采样法(competitive adaptive reweighting sampling, CARS)、移动窗口偏最小二乘法(moving window partial least squares, MWPLS)和蒙特卡洛无信息变量消除法算法对蓝莓红外光谱特征变量的选取影响，通过相关系数(R²)和均方根误差(root mean square error, RMSE)等参数对模型进行评价，得到最优蓝莓含水率近红外预测模型。各特征变量选取算法均能在减少参与建模波长变量基础上，提升模型拟合度及预测能力。其中，CARS-PLS建模方法共选取11个特征变量参与建模，其模型校正相关系数为0.951 0,校正均方根误差为0.042 9,预测相关系数为0.946 5,预测均方根误差为0.047 3。最后建立基于CARS特征变量的蓝莓干燥过程含水率PLS模型，能有效对蓝莓干燥过程含水率进行准确、快...     

基于VIS/NIR高光谱成像技术检测鸡肉嫩度

利用400~1000 nm可见近红外高光谱成像系统对鸡肉嫩度进行快速无损检测研究。采集鸡肉表面的高光谱散射图像,提取样本感兴趣区域反射光谱曲线并用剪切力值表征鸡肉的标准嫩度。以原始光谱和多元散射校正(MSC)预处理光谱数据建立鸡肉嫩度的偏最小二乘回归(PLSR)模型,预处理光谱建立的模型效果更优。基于MSC预处理,采用偏PLS权重系数法结合逐步回归法筛选出了4个特征波长。然后采用PLSR和多元线性回归(MLR)模型分别建立特征波长处光谱反射值和鸡肉嫩度关系的数学模型,优选最佳模型。结果显示:MLR模型预测效果较好,预测相关系数(RP)和均方根误差(RMSEP)分别为0.94和1.97。研究表明:利用可见近红外高光谱成像技术结合多元回归分析法对鸡肉嫩度的快速无损检测是可行的。     

鱼体新鲜度近红外光谱检测方法的比较研究

目的 建立针对淡水鱼整鱼鱼体新鲜度的快速无损检测方法. 方法 通过比较不同的光谱与相应挥发性盐基氮(TVB-N)值的建模结果, 以及对比分析竞争性自适应重加权算法(CARS)、遗传算法(GA)及连续投影算法(SPA)三种特征波长选择方法对模型的优化结果, 对鱼鳞及光谱采集部位等影响因素进行研究。结果 鱼体有鳞时的尾部为最佳新鲜度检测部位。CARS法较优且鱼体新鲜度检测的最优波段为800~1100 nm, 采用CARS特征波长选择方法选择出23个波长变量重新建立PLS模型, 模型预测集相关系数RP=0.957, 预测均方根误差RMSEP=0.589 mg/100 g。利用CARS方法选择的23个波长变量对淡水鱼进行新鲜度评价, 准确率达96.67%。结论 该方法为淡水鱼整鱼新鲜度快速无损检测提供了一种有效的方法。     

基于微型近红外光谱仪的油菜籽含油率模型参数优化研究

为实现油菜籽含油率快速无损检测,采用微型近红外光谱仪,结合竞争性自适应重加权(CARS)、遗传算法(GA)、连续投影算法(SPA)、无信息变量消除法(UVE)、向后区间偏最小二乘法(BIPLS)、联合区间偏最小二乘法(SIPLS)等方法优选油菜籽含油率近红外光谱特征波长,建立偏最小二乘回归(PLSR)和最小二乘支持向量机(LS-SVM)定量分析模型,同时对LS-SVM模型参数进行优化。研究表明,对PLSR模型,BIPLS+GA优选的26个特征波长建模效果最好,其预测相关系数(Rp)和预测均方根误差(RMSEP)分别为0.9330和0.0075,对LS-SVM模型,SIPLS+GA优选的13个特征波长建模效果最好,预测相关系数(Rp)和预测均方根误差(RMSEP)分别0.9192和0.0055。证明了波长优选和参数优化可有效简化油菜籽含油率近红外光谱定量分析模型,提高模型预测精度和稳定性,为进一步拓展微型近红外光谱仪的应用提供技术参考。     

盐池滩羊肉脂肪含量的高光谱预测模型构建

本文利用可见-近红外高光谱成像技术预测冷鲜滩羊肉脂肪含量,优选最佳预测模型。测定90个滩羊背最长肌的脂肪含量并采集其光谱图像,对原始光谱进行不同种预处理后,构建了全波段下的偏最小二乘回归(PLSR)和主成分回归(PCR)的光谱预测模型。为减少模型运算次数,在预处理效果最优的全波段模型上采用连续投影算法(SPA)、应用竞争性自适应重加权(CARS)、变量组合集群分析(VCPA)和波长空间迭代收缩(IVISSA)方法提取特征波长,构建脂肪含量的光谱预测模型。结果表明:采用归一化(Normlize)预处理后光谱构建的PLSR全波段模型效果最好,校正集模型相关系数(Rc)达到0.921;采用多元散射校正(MSC)预处理后光谱构建的PCR全波段模型效果最好,其校正集模型相关系数(Rc)达到0.850;在4种提取特征波长过程中,IVISSA算法所提取特征波长的交互验证均方根误差(RMSECV)最低,为0.0072;Normlize-IVISSA-PLSR模型较其他3种算法所构建的PLSR模型效果最优,其校正集相关系数(Rc)和预测集相关系数(Rp)值分别为0.931和0.754,表明利用高光谱技术对盐池滩羊肉脂肪含量进行预测是可行的。研究成果为冷鲜滩羊肉品质在线光谱快速无损检测系统开发提供理论依据。     

基于高光谱成像技术预测牡蛎干制加工过程中的水分含量

提出一种应用高光谱成像技术结合化学计量学检测牡蛎干制加工过程中水分含量的方法。采用高光谱成像系统,在400～1 100 nm范围内,采集到5个干燥时期的100个牡蛎干样本高光谱图像。提取所有样本感兴趣区域的平均光谱数据,对原始光谱数据进行多元散射校正(MSC)、卷积平滑(S-G)预处理,采用相关系数法提取8个特征波长。基于所提取的特征波长,建立光谱数据与水分含量的多元线形回归(MLR)和BP神经网络模型。结果表明:两种模型均有较好的预测效果。MLR模型的校正集、预测集和交叉验证集的相关系数较BP神经网络低;校正集、预测集和交叉验证集均方根误差分析结果表明,BP神经网络效果较MLR好。高光谱成像技术结合化学计量学方法可检测牡蛎干制过程中水分含量的变化。     

蚁群和遗传算法优化花茶花青素近红外光谱预测模型的比较

以建立花茶花青素含量的最优近红外光谱模型为目标,对比研究了蚁群算法(Ant ColonyOptimization,ACO)和遗传算法(Genetic Algorithm,GA)优化近红外光谱谱区的效果。ACO-i PLS将全光谱划分为12个子区间时,优选出第1、9、10共3个子区间,所建的校正集和预测集相关系数分别为0.901 3和0.864 2;交互验证均方根误差(RMSECV)和预测均方根误差(RMSEP)分别为0.160 0 mg/g和0.202 0 mg/g;GA-i PLS将全光谱划分为15个子区间时,优选出第1、5共2个子区间,所建模型的校正集和预测集相关系数分别为0.906 3和0.879 3,交互验证均方根误差(RMSECV)和预测均方根误差(RMSEP)分别为0.156 0 mg/g和0.206 0 mg/g。研究结果表明:ACO-i PLS和GA-i PLS均可以有效选择近红外光谱特征波长,其中GA-i PLS模型的精度更高。     

蚁群-遗传算法优化近红外光谱检测花茶花青素含量的研究

目的本研究基于蚁群-遗传区间偏最小二乘(ACO-GA-iPLS)近红外谱区筛选方法预测花茶花青素含量。方法首先对花茶近红外光谱进行预处理;然后用ACO-iPLS优选出特征子区间;最后对所选的特征子区间,用GA-iPLS进一步细化花青素的特征子区间,并建立花青素的预测模型。结果优选出3个特征子区间(第1、9、10子区间),所建模型对应的交互验证均方根误差(RMSECV)和预测均方根误差(RMSEP)分别为0.1460mg/g和0.1840 mg/g,校正集和预测集相关系数分别为0.9187和0.8856。结论ACO-GA-iPLS可以有效选择近红外光谱特征波长,简化模型,提高模型精度。     

基于变量筛选的温州蜜桔品质的光谱快速检测

目的利用可见/近红外光谱技术结合变量筛选算法建立预测模型。方法采集7个不同批次蜜桔样本的漫透射光谱,预处理优化后,以无信息变量消除法(uninformative variable elimination,UVE)、竞争性自适应重加权法(competitive adaptive reweighting sampling,CARS)及其组合(UVE-CARS)共3种策略来进行光谱有效波段的筛选,建立蜜桔可溶性固形物含量(soluble solid content,SSC)的偏最小二乘预测模型(partial least square,PLS)。结果比较全变量模型和3个特征变量模型的预测性能,UVE-CARS-PLS模型取得了最优的检测效果,相比全变量模型,建模变量数减少了96.5%,其预测集相关系数R_P提升至0.732,预测集均方根误差(root-mean-square error,RMSEP)下降至0.873~0Brix。结论结合多重变量选择算法,可以进一步压缩建模变量数,简化模型,提高模型预测精度,实现区域蜜桔品质的光谱快速检测。     

速溶普洱茶中水分、咖啡碱和茶多酚含量近红外光谱快速测定方法的建立

以随机抽取的42个市售速溶茶产品为研究对象,采用近红外光谱分析技术并结合偏最小二乘法(PLS)对其水分、咖啡碱和茶多酚含量进行定标建模分析。建模结果以校正集相关系数(R_c)、校正集均方根误差(RMSEC)、交互验证相关系数(R_cv)和交互验证均方根误差(RMSECV)为指标评价模型的优劣。结果表明,水分较为理想的定标模型R_c=0.9266,RMSEC=0.6439,R_cv=0.8809,RMSECV=0.8509;咖啡碱的最优定标模型R_c=0.9964,RMSEC=0.1337,R_cv=0.9543,RMSECV=0.4958;茶多酚较为理想的定标模型R_c=0.9845,RMSEC=1.2097,R_cv=0.9679,RMSECV=1.8083。经验证集样品检验,水分、咖啡碱、茶多酚的预测相关系数分别是0.9050、0.9350、0.9557,与实际测定值吻合度较高,可为速溶普洱茶制品理化成分的快速检测提供参考。     

近红外漫反射光谱法快速检测高纤维素、木质素物料水分含量

目的：建立一种快速检测高纤维素、木质素物料水分含量的方法。方法：以槟榔这种含高纤维素、木质素的中药材为原料,用近红外光谱仪采集近红外漫反射光谱,运用NIR Cal建模软件对光谱数据进行预处理,优选特征波长,并运用偏最小二乘法（PLS）分析建立槟榔水分含量定量模型。结果：槟榔水分含量定量模型校正集决定系数为0.994 2,校正误差均方根（RMSEC）为0.50;验证集决定系数为0.986 7,预测误差均方根（RMSEP）为0.68。结论：该方法简便、快速、安全、实用、准确,适用于含高纤维素、木质素物料的水分含量的快速测定。     

Chemometric Models for the Quantitative Descriptive Sensory Properties of Green Tea (<Emphasis Type="Italic">Camellia sinensis</Emphasis> L.) Using Fourier Transform Near Infrared (FT-NIR) Spectroscopy

Multivariable models based on chemometric analyses of the tea infusion sensory data and FT-NIR spectra of 70 “Biluochun” green tea (Camellia sinensis L.) samples were generated aiming to predict the scores of sensory attributes of green tea. Modified BP_AdaBoost algorithm was used to develop the models. The synergy interval partial least square (siPLS) algorithm was applied to select the wavenumbers for the prediction model of sensory properties in order to take only significant spectral intervals into account. Some parameters were optimized by cross-validation in model calibrations. Experimental results showed that the optimal BP_AdaBoost model was achieved with four principal components (PCs), when 184 variables in the combination of four spectral intervals [3 17 19 21] were selected by siPLS. The predicted precision of the best model obtained were as follows: the root mean square error of cross-validation (RMSECV) was 5.0305 and the correlation coefficient (R _c) was 0.8554 in the calibration set; the root mean square error of prediction (RMSEP) was 6.0807, the correlation coefficient (R _p) was 0.7717, and the ratio performance deviation (RPD) was 1.59 in the prediction set. Finally, the BP_AdaBoost model revealed its superior performance when compared with back propagation neural network (BPNN) model. The overall results demonstrate that FT-NIR spectroscopy technique can be successfully used in the evaluation of sensory quality of green tea, and BP_AdaBoost algorithm shows its superiority in model calibration.     

基于衰减消去蜻蜓算法的小麦粉蛋白质近红外特征波长优选

为优选小麦粉蛋白质近红外光谱特征波长,结合指数和线性衰减函数对单群蜻蜓算法（single-binary dragonflyalgorithm,single-BDA）进行改进并提出一种衰减消去蜻蜓算法（attenuationelimination-BDA,AEBDA）。分别使用single-BDA和AE-BDA筛选160个小麦粉样本中蛋白质近红外光谱的波长,并用偏最小二乘回归法建立蛋白质定量分析模型评价波长选择效果。结果表明：与single-BDA相比,AE-BDA所选波长数量少、稳定性强,建立的模型预测效果最佳,模型最佳的预测决定系数为0.972 7,预测标准偏差为0.281 1。8次AE-BDA实验挑选出特征波长的平均数量为15.8个,占原始波长数的12.6%,其中有3个波长每次均被选中。经近红外光谱解析,各入选的波长均包含在小麦粉蛋白质及背景组分的主要吸收谱带范围内。AE-BDA能够以较高的计算效率从小麦粉近红外光谱中筛选出较少的特征波长建立蛋白质分析模型,提高了模型的预测精度和稳定性,可为近红外分析建模提供一种更加简便有效的波长优选方法。     

Application of Near-Infrared Hyperspectral Imaging with Variable Selection Methods to Determine and Visualize Caffeine Content of Coffee Beans

Hyperspectral imaging covering the spectral range of 874–1734 nm was used to determine caffeine content of coffee beans. Spectral data of 958.24–1628.89 nm were extracted and preprocessed. Partial least squares regression (PLSR) model on the preprocessed full spectra obtained good performance with coefficient of determination of prediction (R ² _p ) of 0.843 and root mean square error of prediction (RMSEP) of 131.904 μg/g. In addition, 10 variable selection methods were applied to select the best optimal wavelengths. The PLSR models on the different optimal wavelengths obtained satisfactory results. The PLSR model on the wavelengths selected by random frog (RF) performed the best, with R ² _p of 0.878 and RMSEP of 116.327 μg/g. The RF wavelength selection combined with the PLSR model also achieved satisfactory visualization of caffeine content between different coffee beans. The overall results indicated that optimal wavelength selection was an efficient method for spectral data preprocessing, and hyperspectral imaging was illustrated as a potential technique for real-time online determination for caffeine content of coffee beans.     

Near-infrared reflectance spectroscopy for the prediction of chemical composition in walnut kernel

In the present work, 116 samples were collected and near-infrared reflectance spectroscopy prediction model for determination of moisture, protein, and fat contents of walnut meal were obtained and evaluated. All the samples were analyzed based on the chemical methods. Meanwhile, they were scanned to obtain their near-infrared reflectance spectrum in the wavelength range of 570–1840 nm. Several preprocess treatments including scattering pretreatments, mathematical pretreatments, and aggression methods were optimized to increase the accuracy of the calibration models according to the coefficient of determination for calibration (R_c²) and the cross-validation (one minus the variance ratio, 1-VR), and the standard error of calibration and cross-validation. The results showed modified partial least square loading was the better aggression method to predict the moisture, proteins, and fats in walnut kernel. The calibration equations obtained indicated that near-infrared reflectance spectroscopy had excellent predictive capacity for the three components with R_c² = 0.965, standard error of calibration = 0.052 for moisture, and R_c² = 0.967, standard error of calibration = 0.191 for proteins, and R_c² = 0.979, standard error of calibration = 0.171 for fats, respectively. The external validation further confirmed the robustness and reliability of the near-infrared reflectance spectroscopy prediction models with the correlation coefficient of actual and predicted values (R²) = 0.952, ratio of performance deviation = 4.14, the standard error of prediction =0.058 for moisture, and R² = 0.977, ratio of performance deviation = 5.55, standard error of prediction = 0.182 for proteins, and R² = 0.990, ratio of performance deviation = 8.64, standard error of prediction = 0.191 for fats, respectively. Near-infrared reflectance spectroscopy is a reliable technology to predict these constituents in walnuts.     

基于高光谱信息的生鲜鸡肉离心损失率快速预测模型构建

本文旨在挖掘900~1700 nm波长范围内的高光谱信息构建生鲜鸡肉离心损失率的快速预测模型。通过采集生鲜鸡肉样品的高光谱图像,并提取图像感兴趣区域的光谱信息,经基线校正(Baseline Correction,BC)、高斯滤波平滑(Gaussian Filter Smoothing,GFS)、多元散射校正(Multiplicative Scatter Correction,MSC)、移动平均值平滑(Moving Average Smoothing,MAS)、中值滤波平滑(Median Filtering Smoothing,MFS)5种光谱预处理后,建立全波段偏最小二乘(Partial Least Squares,PLS)回归模型,并利用回归系数法(Regression Coefficient,RC)、连续投影算法(Successive Projections Algorithm,SPA)和逐步回归法(Stepwise)筛选特征波长,优化全波段模型。结果显示,基于Stepwise法从原始光谱中筛选的16个最优波长(900.6、915.4、1024.0、1089.8、1111.2、1155.6、1165.5、1288.9、1305.4、1433.9、1442.1、1486.7、1493.3、1541.1、1690.1和1693.4 nm)构建的PLS模型预测效果较好,其中,r_C为0.94,RMSEC(Root Mean Square Error of Calibration)为1.43%,r_P为0.94,RMSEP(Root Mean Square Error of Prediction)为1.60%。本文表明,基于高光谱信息构建的PLS模型可快速预测生鲜鸡肉离心损失率。     

灵武长枣蔗糖含量的高光谱无损检测

利用高效液相色谱法检测蔗糖含量,同时运用高光谱成像技术结合化学计量方法建立蔗糖预测模型;通过竞争性自适应加权（competitive adaptive reweighted sampling,CARS）算法、连续投影算法（successive projection algorithm,SPA）和无信息消除变量（uninformative variable elimination,UVE）降维处理,建立特征波段和全波段的主成分回归（principal component regression,PCR）、偏最小二乘回归（partial least squares regression,PLSR）和多元线性回归（multivariable linear regression,MLR）模型。结果表明,采用蒙特卡洛方法剔除异常样本后,相关系数由0.611增大到0.846;正交信号校正法预处理效果最佳,RC和RP分别为0.853和0.794;利用SPA、UVE、CARS、CARS＋SPA和CARS＋UVE五种方法提取了5、21、17、10、18 个特征变量,其中CARS-PCR模型最好,校正集、预测集的相关系数为0.861、0.843,校正集、预测集的均方根误差为0.013 mg/g和0.014 mg/g。综上,高光谱成像技术可以实现长枣蔗糖含量的预测,为更深一步探讨枣的内部品质提供参考。     

Wavelength Selection of Hyperspectral Scattering Image Using New Semi-supervised Affinity Propagation for Prediction of Firmness and Soluble Solid Content in Apples

Hyperspectral scattering image technology is an effective method for nondestructive measurement of internal qualities of agricultural products. However, hyperspectral scattering images contain a large number of redundant data that affect the detection performance and efficiency. A new semi-supervised affinity propagation (AP) (NSAP) algorithm coupled with partial least square regression was proposed to select the feature wavelengths from the hyperspectral scattering profiles of “Golden Delicious” apples for predicting apple firmness and soluble solid content (SSC). Six hundred apples were analyzed in the experiment, 400 of which were used for the calibration model and the remaining 200 apples were used for the prediction model. Compared with full wavelengths, the number of effective wavelengths for apple firmness and SSC prediction selected by NSAP, respectively, decreased to 28 and 40 %. The root mean square error of prediction decreased from 6.6 to 6.1 N and from 0.66 to 0.63 %, respectively, whereas the correlation coefficient increased from 0.840 to 0.862 and from 0.876 to 0.890, respectively. Better prediction accuracy was achieved by the prediction model using selected wavelengths by NSAP than that by traditional AP, SAP, and genetic algorithm. The NSAP approach provided an effective means of wavelength selection using hyperspectral scattering image technique.     

基于近红外高光谱成像技术的涩柿SSC含量无损检测

对150个涩柿采集900~1 700nm波段的近红外高光谱图像信息,利用蒙特卡罗—无信息变量消除(MC-UVE)和连续投影算法(SPA)对感兴趣区域光谱进行波长优选。通过MC-UVE-SPA优选出924.69,928.05,1 112.72,1 270.91,1 365.3,1 402.42,1 453.06,1 547.69nm 8个特征波长,对应的光谱反射率作为柿子可溶性固性物含量(SSC)检测的偏最小二乘回归(PLSR)检测模型输入,其预测集相关系数rpre=0.942,预测集均方根误差RMSEP=1.009°Brix。结果表明,MC-UVE-SPA可以有效提取与柿子SSC含量相关的特征信息,从而保留较少的波长建立较好的预测模型。