手持式黄油桃可溶性固形物可见近红外光谱检测设备研制

桃在鲜果市场中占有重要份额。可溶性固形物含量（soluble solid content,SSC）是衡量桃品质的重要参数,是挑选优质桃以及预测最佳采摘时期的重要决策依据。该研究开发了一款基于可见近红外光谱技术的手持式黄油桃SSC无损检测设备。该设备的硬件系统主要由微型光谱仪、卤素灯、OLED显示屏、微控制器以及自主设计的驱动电路组成。为了评估所开发设备的检测性能,采用北京平谷区种植的黄油桃作为样品进行验证。首先,获取校正集样品在680～940 nm范围内的可见近红外光谱,经5点平均平滑和最大值归一化对光谱预处理建立黄油桃SSC偏最小二乘回归模型并用于预测集样本的SSC分析,预测相关系数和均方根误差分别为0.947和0.728%,单果检测时间不超过2 s。为了提高模型精度和稳定性,将校正集和预测集合并后作为新的校正集进行建模,并将重新构建的模型对独立验证集进行预测,SSC预测值与实测值的相关系数为0.906,均方根误差为0.732%。采用分段直接校正算法将主机模型传递到从机。经过模型传递后,从机对独立验证集SSC的预测值与实测值的相关系数和均方根误差分别为 0.865和0.919%。该手持式SSC检测设备可将SSC预测数据以蓝牙方式传输到手机客户端,借助手机定位功能,在地图上实现黄油桃SSC空间可视化分布。研究结果表明,该手持式SSC无损检测设备可以实现黄油桃SSC的准确测量,借助模型传递算法。实现了模型在不同设备间的有效传递,避免了重复建模,可为该设备批量生产节约大量成本,具有广阔的应用前景。     

基于线性渐变分光近红外光谱仪的苹果可溶性固形物含量无损检测研究

为了探讨线性渐变分光近红外光谱仪在水果内部品质无损快速检测方面应用的可行性及其光谱学机理,本文以基于线性渐变分光近红外光谱仪所采集的苹果光谱数据为自变量,以苹果可溶性固形物含量为因变量,采用偏最小二乘回归结合全交互验证算法,分别对全谱、一倍频、二倍频、三倍频、一二倍频、一三倍频、二三倍频共7个谱区建立校正模型并根据模型准确度选择优化谱区。对优化谱区采用平滑、标准正态变量变换、多元散射校正、一阶导数进行光谱数据预处理优化。最后采用外部验证集对优化预处理的模型进行预测准确度评价。结果显示,基于一二倍频区的校正模型准确度较其他谱区最高,其校正测定系数、校正均方根误差、交互验证测定系数、交互验证均方根误差分别为0.753 9、 0.93、 0.745 9、0.95;在此基础上,多元散射校正预处理后所建模型准确度更高,模型校正测定系数、校正均方根误差、交互验证测定系数、交互验证均方根误差分别为0.755 9、 0.92、 0.748 2、 0.94;外部验证集模型预测测定系数、预测均方根误差分别为0.683 4、 1.01。上述3个模型均通过F检验证明其可溶性固形物含量预测值与参考值之间具有显著的相关关系。结果表明,基于一二倍频区数据所建模型基本可以满足苹果可溶性固形物含量快速无损检测的需求,并可为基于线性渐变分光的便携式近红外光谱仪的应用提供一定的参考。     

牛肉质构特性的近红外光谱无损检测

为了建立基于近红外光谱技术的牛肉质构特性快速检测方法,该试验采集了202个新鲜牛肉样品在800～2500 nm波长范围内的漫反射光谱,测定了牛肉的硬度、弹性、咀嚼性和黏附性,经小波消噪后,分别采用平滑、一阶微分、二阶微分等6种方法预处理,建立了牛肉质构特性的偏最小二乘回归模型,并用最优模型进行预测。结果表明：经小波消噪后采用二阶微分预处理方法建立的牛肉硬度、弹性、咀嚼性的检测模型效果最好,其校正集相关系数 r 均在0.9以上,校正集均方根误差（root means square error of calibration,RMSEC）分别为6.247 N、0.760 mm、14.954 mJ,预测集相关系数均在0.664以上,预测集均方根误差（root means square error of prediction,RMSEP）分别为8.887 N、0.951 mm、22.117 mJ,相对预测误差（ratio of prediction to deviation,RPD）值分别为2.43、1.88、2.32,预测精度较高,能够有效地预测牛肉的硬度、咀嚼性,可以检测精度要求不高的牛肉弹性;试验所建立的牛肉黏附性检测模型的预测性能不是很理想,虽然其校正集和预测集相关系数较高（分别为0.720、0.694）,RMSEC 和 RESEP 均较小（分别为0.302、0.243 N·mm）,但其 RPD 值小于1.5,模型预测精度较差,不可以用于预测未知样品的黏附性,此方法还需进一步研究。研究结果为牛肉质构特性的快速无损评价提供了理论依据。     

农产品产地加工与储藏工程技术分类

生鲜牛肉的含水率对其牛肉的加工、储藏、贸易与食用质量有重要影响,为了提高牛肉的经济价值和食用品质,需要研究牛肉含水率的无损检测技术。以取自不同超市的内蒙小黄牛和鲁西黄牛背最长肌为研究对象,有效样本86个,其中,75%的样本作为校正集,25%的样本作为验证集。采集牛肉新鲜切口处400～1170 nm波长范围内的漫反射光谱,用国标方法测定牛肉含水率。经过多元散射校正（multiplicative scatter correction, MSC）、变量标准化（standard normalized variate, SNV）和直接正交信号校正（direct orthogonal signal correction, DOSC）等方法预处理,在400～1170 nm范围内分别建立多元线性回归（multiple linear regression, MLR）模型、主成分回归（principal component Regression, PCR）模型和偏最小二乘回归（partial least squares regression, PLSR）模型。结果表明使用MSC预处理方法建立的模型预测效果最佳,其中用PLSR建模结果最好,校正集的相关系数和校正标准差分别是0.92和0.0069,验证集的相关系数和验证标准差分别是0.92和0.0047,外部验证的相关系数和验证标准差分别是0.85和0.0054。结果表明,可见/近红外光谱结合MSC预处理方法建立的PLSR模型,可以对牛肉含水率进行准确的快速无损评价,为生鲜牛肉含水率快速无损检测技术的应用提供理论参考。     

基于近红外光谱的室温贮藏下鲜枣霉菌污染动力学模型

为了预测鲜枣常温贮藏的保鲜期,确保鲜枣的品质要求及食用安全,应用近红外光谱建立了室温贮藏下鲜枣内部霉菌菌落总数变化的动力学模型。通过对几种数据预处理方法的比较及特征波数的选择,实现了鲜枣霉菌菌落总数变化的近红外模型的优选。结果表明:经过多元散射校正处理的鲜枣近红外光谱,应用多元线性回归方法建立的霉菌菌落总数模型预测能力较好,校正集相关系数为0.920,均方根误差为1.503,预测集相关系数为0.889,均方根误差为1.514。同时,将近红外光谱模型应用于霉菌菌落总数随贮藏时间变化的零级反应动力学模型中,得到模型的相关系数为0.981。根据近红外光谱吸光度值与贮藏时间的线性关系,当霉菌菌落总数初始值小于等于10cfu/g时,预测出鲜枣在室温下的保鲜期一般为8d。研究表明,结合动力学模型的近红外光谱技术可以作为一种无损、快速检测方法来检测鲜枣霉菌菌落总数变化。     

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。     

可见/近红外光谱技术无损检测果实坚实度的研究

该研究的目的是建立可见/近红外光谱与梨果实坚实度之间的数学模型,评价可见/近红外光谱技术无损测量梨果实坚实度的应用价值.在可见/近红外光谱区域(350～1800nm),试验对比分析了不同测量部位、不同光谱预处理方法和不同校正建模算法的梨果实坚实度校正模型.结果表明:赤道部位吸光度一阶微分光谱的偏最小二乘回归所建梨果实坚实度校正模型的预测性能较优,其校正和预测相关系数分别为0.8779和0.8087,校正和预测均方误差分别为1.0804N和1.4455N.研究表明:可见/近红外光谱技术无损检测梨果实坚实度是可行的.     

基于机器视觉和近红外光谱技术的杏干品质无损检测

干果品质直接影响其市场销售。该研究以杏干为对象探讨用机器视觉和近红外光谱技术快速无损检测干果内外品质的方法。拍摄杏干4个不同位置的彩色图像,用基于区域骨架化的填充法分割杏干,提取每种角度下的面积。从100个正常杏干样本中随机挑选75个为校正集,25个为预测集,用多元线性回归对杏干的实际质量和4个面的面积建模,得到校正集和预测集相关系数分别为0.9374和0.9307,杏干质量分级的准确率为90%。提出用基于平均灰度的区域增长法提取杏干缺陷,缺陷检测的准确率为84.5%。采用SNV对杏干近红外光谱进行预处理,然后分别采用偏最小二乘法(PLS)、向后区间偏最小二乘法(biPLS)及联合区间偏最小二乘法(siPLS)建立杏干糖度预测模型。结果表明,当全光谱范围被划分为22个子区间,优选出区间[17、2、3、9、20、13、7、18、15、11、6],主因子数为10时建立的biPLS糖度模型性能最好。其校正集相关系数和校正均方根误差分别为0.8983和1.23,预测集相关系数和预测均方根误差分别为0.8814和1.46。研究表明,机器视觉结合近红外光谱技术能对杏干内外品质进行综合检测,也可为其他干果的品质检测提供借鉴。     

蛋壳品质的近红外光谱检测分析

蛋壳品质对蛋品孵化、贮存和运输均有重要影响。为了探索近红外光谱技术快速检测蛋壳品质的方法,该文在鸡蛋蛋壳品质指标相关性分析的基础上进行了蛋壳品质的近红外光谱检测分析,研究比较了不同建模方法、不同光谱预处理方法和不同波段范围对预测结果的影响。结果表明：在5段特征波长范围内建立的经过多元散射校正的偏最小二乘回归（PLSR,partial least squares regression）模型对蛋壳强度的预测结果最好,相关系数r为0.86,校正、预测均方根误差分别为4.42、7.53 N;同时蛋壳百分比（蛋壳质量/蛋质量）的PLSR模型的相关系数r为0.92,校正、预测误差分别为0.313%、0.529%;蛋壳厚度的PLSR模型的相关系数r为0.81,校正、预测误差分别为0.0176、0.0234 mm。研究结果表明应用近红外光谱技术预测蛋壳品质是可行的,为蛋壳品质的快速无损检测提供了一种新的方法。     

土壤有机质含量可见-近红外光谱反演模型校正集优选方法

土壤有机质含量可见-近红外光谱反演过程中校正集的构建策略对模型的预测精度有重要影响。以江汉平原洪湖地区水稻土为研究对象,采用Kennard-Stone(KS)法,Rank-KS(RKS)和Sample set Partitioning based on joint X-Y distance(SPXY)法,构建样本数占总校正集不同比例的子校正集,通过偏最小二乘回归,建立土壤有机质含量的可见—近红外光谱反演模型。结果表明:KS法无法提高模型预测精度,但可以在保证标准差与预测均方根误差比(ratio of performance to standard deviation,RPD)2.0的前提下减少30%的校正样本;基于SPXY法的模型,当子校正集样本比例为总校正集的50%时达到最佳的模型预测精度,RPD为2.557;RKS法能够在保证预测精度的情况下(RPD2.0),最多减少总校正集70%的样本,对应模型RPD为2.212。当校正集与验证集的有机质含量分布相近时,能够以较少的建模样本达到与总校正集相近甚至更高的模型预测精度,提升土壤有机质光谱反演模型的实用性。     

基于近红外光谱技术的淡水鱼品种快速鉴别

为探索淡水鱼品种的快速鉴别方法,该文应用近红外光谱分析技术,结合化学计量学方法,对7种淡水鱼品种的判别分类进行了研究。采集了青、草、鲢、鳙、鲤、鲫、鲂等7种淡水鱼,共665个鱼肉样品的近红外光谱数据,经过多元散射校正(multiplicative scatter correction,MSC)、正交信号校正(orthogonal signal correction,OSC)、数据标准化(standardization,S)等20种方法预处理,在1 000~1 799 nm范围内分别采用偏最小二乘法(partial least square,PLS)、主成分分析(principal component analysis,PCA)和BP人工神经网络技术(back propagation artificial neural network,BP-ANN)、偏最小二乘法和BP人工神经网络技术对7种淡水鱼原始光谱数据进行了鉴别分析。结果表明,近红外光谱数据,结合主成分分析和BP人工神经网络技术建立的淡水鱼品种鉴别模型最优,模型的鉴别准确率达96.4%,对未知样本的鉴别准确率达95.5%。模型具有较好的鉴别能力,采用该方法能较为准确、快速地鉴别出淡水鱼的品种。     

基于dbiPLS-SPA变量筛选的固态发酵湿度近红外光谱检测

为了提高基于近红外光谱技术的固态发酵关键过程参数——湿度快速检测的精度和稳定性,研究采用动态反向区间偏最小二乘（dbiPLS）法结合连续投影算法（SPA）进行最佳光谱子区间和特征组合变量的筛选,通过交互验证法确定偏最小二乘（PLS）模型的主成分因子数,并以预测均方根误差（RMSEP）和相关系数（Rp）作为模型的评价标准。试验结果显示,最佳dbiPLS-SPA模型筛选的组合变量个数为8,其RMSEP和Rp分别为1.1795%（质量分数）和0.9430。试验结果表明,dbiPLS-SPA是一个有效的波长组合变量筛选方法,可简化模型结构、增强模型精度和稳健性。     

基于可见/近红外反射光谱的梨树叶片钾含量的快速测定研究

利用可见/近红外反射光谱定量分析技术对梨树鲜叶钾素含量进行快速测定研究。对150个梨树叶片样本进行光谱扫描,其中120个做建模集,30个做验证集。通过对样品的可见/近红外光谱进行多种预处理,并建立钾素预测模型,探讨了可见/近红外光谱数据预处理对预测精度的影响。结果表明,通过原始光谱与S-G(3)平滑相结合的预处理方法,用17个主成分建立的偏最小二乘法模型最好,其交叉验证集和预测集模型的决定系数(R2)分别为0.722 7和0.679 1,交叉验证均方根误差(RMSECV)为1.171,预测的平均相对误差为6.81%,能高效、快速地预测梨树叶片钾素含量,为梨树钾素快速测定提供了新的手段。     

Use of visible and near infrared spectroscopy and least squares-support vector machine to determine soluble solids content and pH of cola beverage

Visible and near infrared (VIS/NIR) transmission spectroscopy and chemometric methods were utilized for the fast determination of soluble solids content (SSC) and pH of cola beverage. A total of 180 samples were used for the calibration set, whereas 60 samples were used for the validation set. Some preprocessing methods were applied before developing the calibration models. Several PLS factors, extracted by partial least squares (PLS) analysis, were used as the inputs of least squares-support vector machine (LS-SVM) model according to their accumulative reliabilities. The correlation coefficient (r), root mean square error of prediction (rmsEP), bias, and RPD were 0.959, 1.136, -0.185, and 3.5 for SSC, whereas 0.973, 0.053, 0.017, and 4.1 for pH, respectively. An excellent prediction precision was achieved by LS-SVM compared with PLS. The results indicated that VIS/NIR spectroscopy combined with LS-SVM could be applied as a rapid and alternative way for the fast determination of SSC and pH of cola beverage.     

Development of a Near‐Infrared Imaging System for Determination of Rice Moisture

The objective of this study was to develop a near‐infrared (NIR) imaging system to determine rice moisture content. The NIR imaging system fitted with 15 band‐pass filters (wavelengths of 870–1,014 nm) was used to capture the spectral image. In this work, calibration methods including multiple linear regression (MLR), partial least squares regression (PLSR), and artificial neural network (ANN) were used in both near‐infrared spectrometry (NIRS) and the NIR imaging system to determine the moisture content of rice. Comprehensive performance comparison among MLR, PLSR, and ANN approaches has been conducted. To reduce repetition and redundancy in the input data and obtain a more accurate network, six significant wavelengths selected by the MLR model, which had high correlation with the moisture content of rice, were used as the input data of the ANN. The performance of the developed system was evaluated through experimental tests for rice moisture content. This study adopted the coefficient of determination (r_val²), the standard error of prediction (SEP), and the relative performance determinant (RPD) as the performance indices of the NIR imaging system with respect to the tests of rice moisture content. Utilizing these three models, the analysis results of r_val², SEP, and RPD for the validation set were within 0.942–0.952, 0.435–0.479%, and 4.2–4.6, respectively. From experimental results, the performance of NIR imaging system was almost the same as that of NIRS. Using the developed NIR imaging system, all of the three different calibration methods (MLR, PLSR, and ANN) provided a high prediction capacity for the determination of moisture in rice samples. These results indicated that the NIR imaging system developed in this study can be used as a device for the measurement of rice moisture content.     

农产品品质近红外光谱分析结果影响因素研究综述

结合国内外研究现状,针对校正集样品选择,样品粒径、含水率、温度、试验因素,以及预处理和数学建模方法的选择对农产品近红外光谱分析结果的影响进行了探讨分析。分析结果表明：为获得可靠的分析结果,样品选择应均匀、广泛,并应考虑样品粒径、含水率及温度的影响,同时应选择合适的光谱预处理和数学建模算法。     

Calibration of visible and near infrared spectroscopy for soil analysis at the field scale on three European farms

The development of accurate calibration models for selected soil properties is a crucial prerequisite for successful implementation of visible and near infrared (Vis‐NIR) spectroscopy for soil analysis. This paper compares the performance of calibration models developed for individual farms with that of general models valid for three farms in three European countries. Fresh soil samples collected from farms in the Czech Republic, Germany and Denmark were scanned with a fibre‐type Vis‐NIR spectrophotometer. After dividing spectra into calibration (70%) and validation (30%) sets, spectra in the calibration set were subjected to partial least squares regression (PLSR) with leave‐one‐out cross‐validation to establish calibration models of soil properties. Except for the Czech Republic farm, individual farm models provided successful calibration for total carbon (TC), total nitrogen (TN) and organic carbon (OC), with coefficients of determination (R²) of 0.85–0.93 and 0.74–0.96 and residual prediction deviations (RPD) of 2.61–3.96 and 2.00–4.95 for the cross‐validation and independent validation respectively. General calibration models gave improved prediction accuracies compared with models of farms in the Czech Republic and Germany, which was attributed to larger ranges in the variation of soil properties in general models compared with those in individual farm models. The results revealed that larger standard deviations (SDs) and wider variation ranges have resulted in larger R² and RPD, but also larger root mean square errors of prediction (RMSEP). Therefore, a compromise solution, which also results in small RMSEP values, should be found when selecting soil samples for Vis‐NIR calibration to cover a wide variation range.     

Rapid prediction of acid detergent fiber, neutral detergent fiber, and acid detergent lignin of rice materials by near-infrared spectroscopy

A rapid predictive method based on near-infrared spectroscopy (NIRS) was developed to measure acid detergent fiber (ADF), neutral detergent fiber (NDF), and acid detergent lignin (ADL) of rice stem materials. A total of 207 samples were divided into two subsets, one subset (approximately 136 samples) for calibration and cross-validation and the other subset for independent external validation to evaluate the calibration equations. Different mathematical treatments were applied to obtain the best calibration and validation results. The highest coefficient of determination for calibration (R2) and coefficient of determination for cross-validation (1-VR) were 0.968 and 0.949 for ADF, 0.846 and 0.812 for NDF, and 0.897 and 0.843 for ADL, respectively. Independent external validation still gave a high coefficient of determination for external validation (r2) and a low standard error of performance (SEP) for the three parameters; the best validation results were SEP = 0.933 and r2 = 0.959 for ADF, SEP = 2.228 and r2 = 0.775 for NDF, and SEP = 0.616 and r2 = 0.847 for ADL, indicating that NIR gave a sufficiently accurate prediction of ADF and ADL content of rice material but a less satisfactory prediction for NDF. This study suggested that routine screening for these forage quality parameters with large numbers of samples is possible with NIRS in early-generation selection in rice-breeding programs.     

Determination of moisture in Cheddar cheese by near infrared reflectance spectroscopy

Near infrared reflectance (NIR) spectroscopy was used to determine the moisture content of Cheddar cheese. Through multiple linear regression analysis, a 3-wavelength calibration was developed for use with a commercial filter monochromator instrument. For a validation set of 47 samples, the correlation coefficient squared (r2) between the NIR and oven moisture methods was 0.92, with a standard error of performance (SEP) of 0.38%. Sample temperature was found to significantly affect the spectral response; therefore, it was necessary to equilibrate all samples to a uniform temperature prior to NIR analysis. Aging may also affect the NIR characteristics of cheese, although it was possible to develop a successful calibration that encompassed a wide range of aging times.     

基于VNIR和机器学习算法的原状土剖面Cu含量预测

快速测量土壤剖面重金属含量是评估土壤重金属污染状况并选择相应修复技术的关键。为了探讨可见光-近红外光谱法（Visible and Near-Infrared Reflectance Spectroscopy,VNIR）预测原状土壤剖面重金属含量的潜力,以江西省两个典型工矿厂周边农田土壤为研究对象,共采集了19个深度约100 cm的完整土壤剖面样品,分别测定土壤剖面样品的VNIR数据及其Cu含量。采用偏最小二乘回归法（Partial Least Squares Regression,PLSR）、Cubist混合线性回归决策树（Cubist Regression Tree,Cubist）、高斯过程回归（Gaussian Process Regression,GPR）和支持向量机（Support Vector Machine Regression,SVM）方法研究不同光谱预处理方法对土壤Cu含量预测精度的影响。结果显示,Cubist、GPR和SVM这三种机器学习算法的预测精度普遍高于PLSR,其中一阶导数（First-Order Derivative,FD）预处理的SVM模型预测精度最高（R2=0.95,均方根误差为7.94 mg/kg,相对分析误差为4.34）。这表明利用VNIR和机器学习可以对原状土壤剖面Cu含量进行有效预测,为快速监测Cu及其他重金属含量的相关研究提供参考。