近红外光谱联合CARS-PLS-LDA的山茶油检测

为了寻找快速判别山茶油掺假的检测方法,本研究利用近红外光谱技术对掺杂大豆没油山茶油进行掺假检测研究.试验在350～1 800 nm波段范围内采集样本的透射光谱,利用CARS方法筛选重要的波长变量,应用偏最小二乘-线性判别分析(PLS-LDA)建立山茶油掺假的判别模型,并与未经变量优选的判别模型进行比较.结果表明,近红外光谱技术联合CARS-PLS-LDA方法可以有效判别纯山茶油和掺假山茶油,校正集、预测集及独立样本组样本的判别正确率、灵敏度及特异性均为100％.CARS-PLS-LDA判别模型性能优于未经变量优选的判别模型,表明CARS方法可以有效筛选重要波长变量,能简化判别模型及提高判别模型的稳定性和判别精度.本研究可为山茶油掺假快速检测提供理论依据.     

柿子可溶性固形物含量的可见-近红外光谱检测

为了实现柿子（Diospyros kaki thunb）可溶性固形物含量的快速无损检测,提出了一种采用可见-近红外光谱分析技术无损检测柿子可溶性固形物含量的方法。采用Field Spec 3光谱仪对3种不同品种的柿子进行光谱分析,共获取66个样本数据。利用平均平滑法对样本数据进行预处理,再采用主成分分析法,依据可信度获取光谱的6个主成分数据。将样本随机分成51个建模样本（每种各17个）和15个验证样本（每种各5个）,把6个主成分数据作为BP神经网络的输入变量,柿子的可溶性固形物含量作为输出变量,隐含层的节点数为11,建立3层BP神经网络检测模型,并用该模型对15个验证样本进行预测。结果表明,所建校正模型的校正标准差（SEC）为0.232,对预测集样本可溶性固形物含量的预测相对误差在3%以下,预测值和实测值的决定系数（R2）为0.99,预测标准差（SEP）为0.257。结果表明应用近红外光谱技术结合主成分分析和神经网络算法检测柿子的可溶性固形物含量是可行的。     

基于小波变换的番茄总糖近红外无损检测

分别采用小波消噪、常数偏移消除等11种光谱预处理方法,对番茄总糖含量（质量分数）的近红外光谱进行预处理,通过偏最小二乘法定量校正模型预测值比较得出,小波消噪是适合番茄近红外光谱的最佳预处理方法,小波消噪的总糖质量分数近红外光谱优选区域为11 998.9～6 097.8 cm-1和4 601.3～4 246.5 cm-1,在此光谱区内建立的番茄总糖质量分数偏最小二乘法模型预测值与实测值的相关系数为0.930,内部交叉验证均方差为0.466%,校正标准差为0.469%,预测标准差为0.260%。试验结果表明：小波消噪后建立的近红外光谱模型能准确地对番茄总糖含量进行快速无损检测。     

便携式番茄多品质参数可见/近红外检测装置研发

该文针对番茄独特的囊室结构及整体成熟度不均等问题,基于可见/近红外全透射光谱,研发了便携式番茄内外品质快速无损实时检测装置.该装置的硬件系统主要包括光源模块、信号采集模块、信号处理模块、电源模块、散热模块和打印模块,基于该硬件系统,采集了番茄630~1100 nm范围内可见/近红外透射光谱,选取650~1100 nm范围的光谱进行SG卷积平滑(savitzky-Golay smooth,SG-smooth)、标准正态变量变换(standard normal variable transformation,SNV)和多元散射校正(muliplication scattering correction,MSC)等预处理,建立了番茄颜色、硬度、总酸、总糖含量的偏最小二乘预测模型.基于QT开发框架编写了番茄多品质无损检测实时分析控制软件,植入番茄多品质参数预测模型,实现了番茄多品质参数检测一键式操作.为了测试该装置的检测精度和稳定性,选取与建模无关的20个同品种样品对每个样品的内外品质重复检测8次,结果表明:番茄颜色预测值与实测值相关系数为0.9528,均方根误差为2.7038,硬度预测值与实测值相关系数为0.9405,均方根误差为0.4486 kg/cm2,总酸含量的预测值与实测值相关系数为0.9537,均方根误差为0.3263%,总糖含量预测值与实测值相关系数为0.9610,均方根误差为0.1974%.番茄样品颜色、硬度、总酸和总糖重复检测最大相对误差分别为2.9%、1.9%、2.0%和1.6%.该便携式检测装置基于可见近红外全透射光谱,实现了番茄颜色、硬度、总酸、总糖含量的同时快速无损实时检测,预测精度及稳定性较好,可以满足实时评价番茄品质的市场需求.     

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。     

基于近红外光谱技术的蜂蜜掺假识别

为了实现蜂蜜掺假的快速识别,应用近红外光谱结合模式识别方法对蜂蜜掺假现象进行了识别分析。该研究收集了中国不同品种、不同地域的典型天然蜂蜜样品,根据目前市场上常见的蜂蜜掺假手段,掺假物质及相对含量情况配制了掺假蜂蜜样品,利用傅立叶近红外光谱仪采集其透反射近红外光谱,分别采用偏最小二乘判别分析（PLS-DA）,独立软模式法（SIMCA）,误差反向传播神经网络（BP-ANN）和最小二乘支持向量机（LS-SVM）等模式识别方法,进行蜂蜜掺假识别研究。研究结果表明：利用这4种方法在蜂蜜中掺入果葡糖浆和果葡糖水的情况下均能很好地识别出掺假蜂蜜样品,其中对于掺入果葡糖浆的掺假情况,校正集的正确判别率均达到95%以上,验证集的正确判别率均达到87%以上,对于掺入果葡糖水的掺假蜂蜜校正集的正确判别率均达到93%以上,验证集的正确判别率均达到84%以上。通过比较4种不同的识别算法,发现采用LS-SVM时,对两种掺假情况下校正集和验证集的正确判别率均达到了100%,表明基于近红外光谱的蜂蜜掺假快速准确识别是可行的。     

果蔬品质手持式近红外光谱检测系统设计与试验

为满足果蔬加工过程快速检测和质量控制的实际需求,研发近红外光谱技术的低成本、实用化、小型化的果蔬品质手持式检测系统。在分析当前近红外光谱实用化过程的瓶颈问题的基础上,提出果蔬品质的手持式检测系统设计方案,阐述了硬件系统选择和软件系统构建,介绍了检测系统的工作原理;选用近红外微机电系统的数字微镜器件作为分光元件,以单点探测器获取检测信息,从而实现光谱检测系统的微型化设计和系统成本的显著降低。以检测番茄为例,利用设计的手持式检测系统,获取番茄900~1 700 nm范围的近红外光谱,利用先选择特征波段再优选波长的建模策略,分别建立了番茄中番茄红素和可溶性固形物含量的定量检测模型;可溶性固形物含量模型的预测相关系数和预测均方根误差分别为0.899和0.133%;番茄红素模型的预测相关系数和预测均方根误差分别为0.886和2.508 mg/kg。研究表明该系统能够满足果蔬品质的快速无损检测要求,可为实用化、小型化的手持式光谱检测仪设计和开发提供参考。     

基于近红外光谱的脐橙产地溯源研究

为研究近红外光谱分析技术鉴别脐橙产地的可行性,该文采用江西、重庆和湖南3个产地脐橙样品1140～1170nm波段的近红外光谱经一阶导数(9点平滑)预处理,分别建立了簇类独立软模式法脐橙产地鉴别模型。在5%显著水平下,模型对3个产地训练集样品的识别率均为100%,拒绝率分别为85.7%、83.3%、100%;对验证集样品的识别率均为100%,拒绝率分别为100%、89.5%、100%,表明簇类独立软模式法模型基本能够判别脐橙产地。将江西、重庆和湖南3个产地的脐橙样品分别赋值0、1、?1,在全波段范围内建立原始光谱脐橙产地的偏最小二乘判别模型,其预测值与真实值的决定系数为0.973,校正标准差为0.110,预测标准差为0.159,模型对训练集和验证集样品的识别率达到100%。因此,应用近红外光谱分析技术可准确、快速地追溯脐橙产地来源。     

基于近红外光谱技术的灵芝提取物掺假定量快速无损检测研究

为了探索快速测定灵芝提取物掺假(淀粉)含量的方法,采用近红外光谱扫描掺杂0%、5%、10%、20%、40%、60%、80%淀粉的灵芝提取物,对其光谱进行预处理和波段选择,并结合偏最小二乘法(PLS)建立灵芝提取物掺假定量快速无损检测方法。结果表明,使用多元散射校正(MSC)预处理方法,波数范围8 000~7 500、6 000~5 500和5 000~4 000cm~(-1),主因子数为8时,建立的偏最小二乘法模型的校正决定系数(R_(cal))、校正均方根差(RMSECV)、验证决定系数(R_(val))和验证均方根差(RMSEP)分别为0.9962、0.0249、0.9960和0.0241。运用该模型对验证集样品进行预测并统计分析,可知预测值与实际掺假值之间无显著差异。本研究为灵芝功能食品市场检测提供了方法基础。     

近红外技术快速测定肉鸡粪便主要肥料成分含量的研究

该文探讨了利用近红外光谱分析技术(NIRS)快速测定肉鸡粪便主要肥料成分含量的可行性。在饲养试验过程中采集了肉鸡粪便样品183个，利用常规实验室分析方法测定了其中的总氮(TN)、总磷(TP)、总钾(TK)、铵态氮(AN)、有效磷(EP)含量，利用近红外光谱仪取得了样品在1421.5～2572.2 nm波段的光谱，并使用偏最小二乘法(PLS)结合交互验证(CV)和留一检验法(LOO)建立了肉鸡粪便肥料成分的近红外光谱定标模型，同时利用该模型对20个样品的总氮、总磷、总钾、铵态氮、有效磷的含量进行预测，得出的5种主要肥料成分的预测值和真实值(实验室经典化学分析方法测定值)之间具有显著的相关性，其相关系数分别为0.9574，0.9590，0.9870，0.9572和0.9650。预测标准差分别为0.0014，0.0012，0.0012，2.3041×10^-4和0.3814。结果表明，利用近红外光谱法对肉鸡粪便风干样品的主要肥料成分进行快速测定是可行的。     

Potential of SPME-GC and chemometrics to detect adulteration of soft fruit purées

The potential of combining solid-phase microextraction (SPME) with gas chromatography and chemometric data analysis to differentiate between pure strawberry samples (Fragraria ananassa) and strawberry samples adulterated with 10, 40, and 70% (v/v) apple purée was investigated. The method involved the extraction of aroma volatiles from the headspace of the purée samples using a SPME fiber followed by GC analysis with flame ionization detection. The principal component analysis (PCA) data matrix consisted of the relative percent peak areas of 37 compounds deemed to be significant in the differentiation of the samples on the basis of adulteration. The PCA results clearly showed that differentiation of the adulterated and unadulterated samples was possible, particularly at the higher levels of adulteration. Partial least-squares regression (PLSR) using a dummy set of Y variables (set to 0 for unadulterated and 1 for adulterated samples) resulted in clear discrimination between unadulterated purées and those containing 40 and 70% (v/v) apple. PLSR using a second set of Y variables, consisting of the actual level of adulteration, enabled quantification of apple purée with a standard error of prediction of 11.6%, implying a minimum detectable level of 25% (v/v) apple. GC-MS analysis enabled identification of the compounds with the greatest influence on sample differentiation. These compounds were identified as hexanoic acid, 2-hexenal, and alpha-farnesene, all of which are key aroma compounds in apples.     

羊肉纯度电子舌快速检测方法

为实现掺假羊肉的快速、客观评价,利用电子舌对混入不同比例鸡肉的掺假羊肉糜进行检测及定性和定量分析。3种浸提溶液分别浸提,样品量均对电子舌传感器的响应影响极显著;以数据点重复性和聚类效果为依据,采用主成分分析方法确定了电子舌检测羊肉糜样品的较佳条件为0.1 mol/L KCl溶液浸提15 g肉糜样品。在此较佳条件下,对混入不同比例鸡肉的掺假羊肉进行检测,结果表明：采用主成分分析和典则判别分析,前2个主成分累积贡献率均超过80%,电子舌均能很好地区分混入不同比例鸡肉的羊肉糜样品;采用多元线性回归分析和偏最小二乘回归分析建立的定量预测模型能有效预测混入的鸡肉比例（R2>0.99,RMSE<3%）。试验表明：电子舌在羊肉掺入鸡肉的鉴别中具有可行性,研究结果可为羊肉掺假鉴别提供参考。     

二维相关光谱结合偏最小二乘法测定牛奶中的掺杂尿素

为了检验牛奶中是否掺杂尿素并将其量化测定,配置含有尿素质量浓度范围为1～20g/L之间40个牛奶样品,以掺杂物尿素浓度为外扰,分别研究了掺杂尿素牛奶的二维相关(近红外-近红外,中红外-中红外,近红外-中红外)光谱特性,在此基础上,分别选择随浓度变化大的4200～4800cm-1和1400～1704cm-1为建模区间,采用偏最小二乘方法建立定量分析模型。研究结果表明:4200～4800cm-1建模分析效果优于1400～1704cm-1建模结果,其交叉验证均方根误差为0.266g/L,对未知样品集预测相关系数达到0.999,预测均方根误差为0.219g/L,这表明所建模型具有较好的预测效果。该方法无需样品处理,成本低,为快速判别牛奶是否掺杂提供了一种新的可能的方法。     

基于生物散斑图像和惯性矩谱分析的牛肉掺腐检测

牛肉掺假严重危害消费者的健康与经济利益,因此对牛肉掺假进行无损检测具有重要意义。该文基于生物散斑技术对牛肉掺假进行定量检测。试验将新鲜牛肉和非新鲜牛肉按不同比例(0、1%、3%、5%~60%(5%梯度)和100%)混合制备掺假样本,并采集样本的生物散斑图像。针对单列惯性矩(inertia moment,IM)表征样本生物活性存在稳定性差的问题,首次提出惯性矩谱(IM谱)分析的方法并用于建立基于支持向量回归机(support vector regression machine,SVR)的牛肉掺假检测模型。结果表明基于IM谱建立的SVR模型能较为准确预测牛肉中掺假物含量,校正集和测试集的决定系数分别为0.85和0.81,均方根误差分别为0.12和0.11。该研究证明了利用生物散斑技术和惯性矩谱分析方法对新鲜牛肉中掺杂腐败牛肉进行定量检测是可行的。     

Updating a synchronous fluorescence spectroscopic virgin olive oil adulteration calibration to a new geographical region

Detecting and quantifying extra virgin olive adulteration is of great importance to the olive oil industry. Many spectroscopic methods in conjunction with multivariate analysis have been used to solve these issues. However, successes to date are limited as calibration models are built to a specific set of geographical regions, growing seasons, cultivars, and oil extraction methods (the composite primary condition). Samples from new geographical regions, growing seasons, etc. (secondary conditions) are not always correctly predicted by the primary model due to different olive oil and/or adulterant compositions stemming from secondary conditions not matching the primary conditions. Three Tikhonov regularization (TR) variants are used in this paper to allow adulterant (sunflower oil) concentration predictions in samples from geographical regions not part of the original primary calibration domain. Of the three TR variants, ridge regression with an additional 2-norm penalty provides the smallest validation sample prediction errors. Although the paper reports on using TR for model updating to predict adulterant oil concentration, the methods should also be applicable to updating models distinguishing adulterated samples from pure extra virgin olive oil. Additionally, the approaches are general and can be used with other spectroscopic methods and adulterants as well as with other agriculture products.     

Quantitative analysis and detection of adulteration in crab meat using visible and near-infrared spectroscopy

Visible and near-infrared spectroscopy (VIS/NIR) has been used to detect economic adulteration of crab meat samples. Atlantic blue and blue swimmer crab meat samples were adulterated with surimi-based imitation crab meat in 10% increments. Waveform evaluation revealed that the main features seen in the spectral data arise from water absorptions with a decrease in sample absorbance with increasing adulteration level. Prediction and quantitative analysis was done using raw data, a 15-point smoothing average, a first derivative, a second derivative, and 150 wavelength spectral data gathered from a correlogram. Regression analysis included partial least squares (PLS) and principal component analysis (PCR). Both models were able to perform similarly in predicting crab meat adulteration. The best model for both PLS and PCR used the first derivative spectral data gathered from the correlogram, with a standard error of prediction (SEP) of 0.252 and 0.244, respectively. The results suggest that VIS/NIR technology can be successfully used to detect adulteration in crab meat samples adulterated with surimi-based imitation crab meat.     

Near-infrared spectroscopic technique for detection of beef hamburger adulteration

A near-infrared spectroscopic technique was developed to detect beef hamburgers adulterated with 5-25% mutton, pork, skim milk powder, or wheat flour with an accuracy up to 92.7%. The accuracy of detection increased with the increase of adulteration level. When an adulterant was detected, the adulteration level was further predicted by calibration equations. The established calibration equations for predicting adulteration levels with mutton, pork, skim milk powder, and wheat flour had standard errors of cross-validation of 3.33, 2.99, 0.92, and 0.57% and coefficients of variance of 0.87, 0.89, 0.99, and 1.00, respectively. The results of this study indicate that near-infrared spectroscopy is potentially useful in detection of beef hamburger adulteration.     

Detection and quantification of species authenticity and adulteration in crabmeat using visible and near-infrared spectroscopy

Seafood processing often removes morphological properties of seafood species that enable the consumer to distinguish one type of organism from another. For this reason, species substitution is the most common form of economic adulteration in the seafood industry. Visible and near-infrared spectroscopy (Vis/NIR) has been used to detect and quantify species authenticity and adulteration in crabmeat samples. Atlantic blue crabmeat was adulterated with blue swimmer crabmeat in 10% increments. Water absorption bands dominated the main features in the crabmeat spectra, with a decrease in sample absorbance with increasing adulteration percentage. Several data pretreatments, i.e., moving average, combing, first and second derivatives, and multiplicative scatter correction, in addition to the raw data, were investigated for prediction and quantitative data analysis using partial least-squares. In addition, quantitative analysis was done using the full spectrum and a sequential approach in which 50 wavelengths were added sequentially to determine a new model and find an optimal solution. The results suggest that Vis/NIR spectroscopy is a suitable technology that can be applied to detect and quantify species authenticity and adulteration in crabmeat.     

高光谱图像结合特征变量筛选定量检测羊肉中狐狸肉掺假

为了探讨快速无损检测羊肉糜中狐狸肉掺假含量的可行性,该研究利用高光谱技术结合特征变量筛选方法开展了其定量检测研究。利用遗传算法、竞争性自适应重加权算法和二维相关光谱分析（Two-Dimensional Correlation Spectroscopy,2D-COS）3种方法分别对代表性样品全部846个波长进行特征波长筛选,得到207、34和14个特征波长;基于全部波长和特征波长建立羊肉糜中狐狸肉掺假含量的偏最小二乘回归（Partial Least Squares Regression,PLSR）和支持向量回归（Support Vector Regression,SVR）模型并进行比较。研究结果表明,基于全部波长和特征波长建立的SVR模型性能均优于PLSR模型。其中,利用2D-COS方法提取的14个特征波长建立的SVR模型（即2D-COS-SVR模型）性能最优,其预测集决定系数和均方根误差分别为0.928和3.00%,相对分析误差为4.85,表明高光谱结合2D-COS-SVR模型可以有效实现羊肉糜中狐狸肉掺假的定量检测。该研究结果为开发低成本肉类掺假检测系统提供技术支持和参考依据。