Chemometric Models for the Quantitative Descriptive Sensory Properties of Green Tea (Camellia sinensis L.) Using Fourier Transform Near Infrared (FT-NIR) Spectroscopy

Near-Infrared Spectroscopy Combined with Fuzzy Improved Direct Linear Discriminant Analysis for Nondestructive Discrimination of Chrysanthemum Tea Varieties

The quality of chrysanthemum tea has a great connection with its variety. Different types of chrysanthemum tea have very different efficacies and functions. Moreover, the discrimination of chrysanthemum tea varieties is a significant issue in the tea industry. Therefore, to correctly and non-destructively categorize chrysanthemum tea samples, this study attempted to design a novel feature extraction method based on the fuzzy set theory and improved direct linear discriminant analysis (IDLDA), called fuzzy IDLDA (FIDLDA), for extracting the discriminant features from the near-infrared (NIR) spectral data of chrysanthemum tea. To start with, a portable NIR spectrometer was used to collect NIR data for five varieties of chrysanthemum tea, totaling 400 samples. Secondly, the raw NIR spectra were processed by four different pretreatment methods to reduce noise and redundant data. Thirdly, NIR data dimensionality reduction was performed by principal component analysis (PCA). Fourthly, feature extraction from the NIR spectra was performed by linear discriminant analysis (LDA), IDLDA, and FIDLDA. Finally, the K-nearest neighbor (KNN) algorithm was applied to evaluate the classification accuracy of the discrimination system. The experimental results show that the discrimination accuracies of LDA, IDLDA, and FIDLDA could reach 87.2%, 94.4%, and 99.2%, respectively. Therefore, the combination of near-infrared spectroscopy and FIDLDA has great application potential and prospects in the field of nondestructive discrimination of chrysanthemum tea varieties.

Deep Learning and Hyperspectral Images Based Tomato Soluble Solids Content and Firmness Estimation

Cherry tomato (Solanum lycopersicum) is popular with consumers over the world due to its special flavor. Soluble solids content (SSC) and firmness are two key metrics for evaluating the product qualities. In this work, we develop non-destructive testing techniques for SSC and fruit firmness based on hyperspectral images and the corresponding deep learning regression model. Hyperspectral reflectance images of over 200 tomato fruits are derived with the spectrum ranging from 400 to 1,000 nm. The acquired hyperspectral images are corrected and the spectral information are extracted. A novel one-dimensional (1D) convolutional ResNet (Con1dResNet) based regression model is proposed and compared with the state of art techniques. Experimental results show that, with a relatively large number of samples our technique is 26.4% better than state of art technique for SSC and 33.7% for firmness. The results of this study indicate the application potential of hyperspectral imaging technique in the SSC and firmness detection, which provides a new option for non-destructive testing of cherry tomato fruit quality in the future.

Discrimination of the Red Jujube Varieties Using a Portable NIR Spectrometer and Fuzzy Improved Linear Discriminant Analysis

In order to quickly, nondestructively, and effectively distinguish red jujube varieties, based on the combination of fuzzy theory and improved LDA (iLDA), fuzzy improved linear discriminant analysis (FiLDA) algorithm was proposed to classify near-infrared reflectance (NIR) spectra of red jujube samples. FiLDA shows performs better than iLDA in dealing with NIR spectra containing noise. Firstly, the portable NIR spectrometer was employed to gather the NIR spectra of five kinds of red jujube, and the initial NIR spectra were pretreated by standard normal variate transformation (SNV), multiplicative scatter correction (MSC), Savitzky-Golay smoothing (S-G smoothing), mean centering (MC) and Savitzky-Golay filter (S-G filter). Secondly, the high-dimensional spectra were processed for dimension reduction by principal component analysis (PCA). Then, linear discriminant analysis (LDA), iLDA and FiLDA were applied to extract features from the NIR spectra, respectively. Finally, K nearest neighbor (KNN) served as a classifier for the classification of red jujube samples. The highest classification accuracy of this identification system for red jujube, by using FiLDA and KNN, was 94.4%. These results indicated that FiLDA combined with NIR spectroscopy was an available method for identifying the red jujube varieties and this method has wide application prospects.

Determination of ethanol content during simultaneous saccharification and fermentation (SSF) of cassava based on a colorimetric sensor technique

Ethanol content is an important indicator reflecting the yield of simultaneous saccharification and fermentation (SSF) of cassava. This study proposes an innovative method based on a colorimetric sensor technique to determine the ethanol content during the SSF of cassava. First, 14 kinds of porphyrin material and one kind of pH indicator were used to form a colorimetric sensor array for collecting odor data during the SSF of cassava. Then, the ant colony algorithm (ACO) and the simulated annealing algorithm (SA) were used to optimize and reconstruct the input color feature components of the support vector regression (SVR) model. The differential evolution algorithm (DE) was used to optimize the penalty factor (c) and the kernel function (g) of the SVR model. The results obtained showed that the combined prediction model of SA-DE-SVR had the highest accuracy, and the coefficient of determination (R P 2) in the prediction set was 0.9549, and the root mean square error of prediction (RMSEP) was 0.1562. The overall results reveal that the use of a colorimetric sensor technique combined with different intelligent optimization algorithms to establish a model can quantitatively determine the ethanol content in the SSF of cassava, and has broad development prospects.

Prediction of specialty coffee flavors based on near-infrared spectra using machine- and deep-learning methods

BACKGROUND

Specialty coffee fascinates people with its bountiful flavors. Currently, flavor descriptions of specialty coffee beans are only offered by certified coffee cuppers. However, such professionals are rare, and the market demand is tremendous. The hypothesis of this study was to investigate the feasibility to train machine learning (ML) and deep learning (DL) models for predicting the flavors of specialty coffee using near-infrared spectra of ground coffee as the input. Successful model development would provide a new and objective framework to predict complex flavors in food and beverage products.

RESULTS

In predicting seven categories of coffee flavors, the models developed using the ML method (i.e. support vector machine) and the deep convolutional neural network (DCNN) achieved similar performance, with the recall and accuracy being 70-73% and 75-77% respectively. Through the proposed visualization method - a focusing plot - the potential correlation among the highly weighted spectral region of the DCNN model, the predicted flavor categories, and the corresponding chemical composition are presented.

CONCLUSION

This study has proven the feasibility of applying ML and DL methods on the near-infrared spectra of ground coffee to predict specialty coffee flavors. The effective models provided moderate prediction for seven flavor categories based on 266 samples. The results of classification and visualization indicate that the DCNN model developed is a promising and explainable method for coffee flavor prediction. © 2021 Society of Chemical Industry.

Determination of acid value during edible oil storage using a portable NIR spectroscopy system combined with variable selection algorithms based on an MPA-based strategy

BACKGROUND

The acid value is an important indicator for evaluating the quality of edible oil during storage. This study employs a portable near-infrared (NIR) spectroscopy system to determine the acid value during edible oil storage. Four MPA-based variable selection methods, namely competitive adaptive reweighted sampling (CARS), the variable iterative space shrinkage approach (VISSA), iteratively variable subset optimization (IVSO), and bootstrapping soft shrinkage (BOSS) were introduced to optimize the preprocessed NIR spectra. Support vector machine (SVM) models based on characteristic spectra obtained by different selection methods were then established to achieve quantitative detection of the acid value during edible oil storage.

RESULTS

The results revealed that, compared with the full-spectrum SVM model, the SVM models established by the characteristic wavelengths optimized by the variable selection methods based on the MPA strategy exhibit a significant improvement in complexity and generalization performance. Furthermore, compared with the CARS, VISSA, and IVSO methods, the BOSS method obtained the least number of characteristic wavelength variables, and the SVM model established based on the optimized features of this method exhibited the optimal prediction performance. The root mean square error of prediction (RMSEP) was 0.11 mg g-1, the coefficient of determination (Rp2) was 0.92 and the ratio performance deviation (RPD) was 2.82, respectively.

CONCLUSION

The overall results indicate that the variable selection methods based on the MPA strategy can select more targeted characteristic variables. This has good application prospects in NIR spectra feature optimization. © 2020 Society of Chemical Industry.

Determination of Fatty Acid Content of Rice during Storage Based on Feature Fusion of Olfactory Visualization Sensor Data and Near-Infrared Spectra

This study innovatively proposes a feature fusion technique to determine fatty acid content during rice storage. Firstly, a self-developed olfactory visualization sensor was used to capture the odor information of rice samples at different storage periods and a portable spectroscopy system was employed to collect the near-infrared (NIR) spectra during rice storage. Then, principal component analysis (PCA) was performed on the pre-processed olfactory visualization sensor data and the NIR spectra, and the number of the best principal components (PCs) based on the single technique model was optimized during the backpropagation neural network (BPNN) modeling. Finally, the optimal PCs were fused at the feature level, and a BPNN detection model based on the fusion feature was established to achieve rapid measurement of fatty acid content during rice storage. The experimental results showed that the best BPNN model based on the fusion feature had a good predictive performance where the correlation coefficient (R_P) was 0.9265, and the root mean square error (RMSEP) was 1.1005 mg/100 g. The overall results demonstrate that the detection accuracy and generalization performance of the feature fusion model are an improvement on the single-technique data model; and the results of this study can provide a new technical method for high-precision monitoring of grain storage quality.

Evaluating aroma quality of black tea by an olfactory visualization system: Selection of feature sensor using particle swarm optimization

Aroma is an important index to evaluate the quality and grade of black tea. This work innovatively proposed the sensory evaluation of black tea aroma quality based on an olfactory visual sensor system. Firstly, the olfactory visualization system, which can visually represent the aroma quality of black tea, was assembled using a lab-made color sensitive sensor array including eleven porphyrins and one pH indicator for data acquisition and color components extraction. Then, the color components from different color sensitive spots were optimized using the particle swarm optimization (PSO) algorithm. Finally, the back propagation neural network (BPNN) model was developed using the optimized characteristic color components for the sensory evaluation of black tea aroma quality. Results demonstrated that the BPNN models, which were developed using three color components from FTPPFeCl (component G), MTPPTE (component B) and BTB (component B), can get better results based on comprehensive consideration of the generalization performance of the model and the fabrication cost of the sensor. In the validation set, the average of correlation coefficient (R_P) value was 0.8843 and the variance was 0.0362. The average of root mean square error of prediction (RMSEP) was 0.3811 and the variance was 0.0525. The overall results sufficiently reveal that the optimized sensor array has promising applications for the sensory evaluation of black tea products in the process of practical production.

Comparison of algorithms for wavelength variables selection from near-infrared (NIR) spectra for quantitative monitoring of yeast (Saccharomyces cerevisiae) cultivations

Rapid monitoring with near-infrared (NIR) spectroscopy of Saccharomyces cerevisiae cultivations was implemented to monitor yeast concentrations. The measurement of one spectrum by using of FT-NIR spectrometer can obtain 1557 wavelength variables. To distinguish which wavelength variables of the collected FT-NIR spectra carry important and relevant information regarding the yeast concentrations, there are three different variables selection approaches, namely genetic algorithm (GA), competitive adaptive reweighted sampling (CARS), and variable combination population analysis (VCPA), were compared in this study. The selected wavelength variables from each method were evaluated using partial least squares (PLS) models to seek the most significant variable combinations for predicting the yeast concentrations. Experimental results showed that the VCPA-PLS model with the best predictive performance was found when using ten principal components (PCs) based on selected eleven characteristic wavelength variables by VCPA algorithm. And the predictive performance indicators of the model were as follows: the root mean square error of prediction (RMSEP) was 0.0680, the coefficient of determination (R_p²) was 0.9924, and the ratio performance deviation (RPD) was 11.8625 in the validation process. Based on the results, it is promising to develop a specific inexpensive NIR sensor for the yeast cultivation process using several light-emitting diodes.

Rapid and non-destructive discrimination of special-grade flat green tea using Near-infrared spectroscopy

Special-grade green tea is a premium tea product with the best rank and high value. Special-grade green tea is normally classified by panel sensory evaluation which is time and sample costly. Near-infrared spectroscopy is considered as a promising rapid and non-destructive analytical technique for food quality evaluation and grading. This study established a discrimination method of special-grade flat green tea using Near-infrared spectroscopy. Full spectrum was used for partial least squares (PLS) modelling to predict the sensory scores of green tea, while specific spectral regions were used for synergy interval-partial least squares (siPLS) modelling. The best performance was achieved by the siPLS model of MSC + Mean Centering pretreatments and subintervals from 15 intervals. The optimal model was used to discriminate special-grade flat green tea with the prediction accuracy of 97% and 93% in the cross-validation and external validation respectively. The chemical compositions of green tea samples were also analyzed, including polyphenols (total polyphenols, catechins and flavonol glycosides), alkaloids and amino acids. Principal components analysis result showed that there is potential correlation between specific spectral regions and the presence of polyphenols and alkaloids. Thus, NIR technique is a practical method for rapid and non-destructive discrimination of special-grade flat green tea with chemical support.

Intelligent evaluation of color sensory quality of black tea by visible-near infrared spectroscopy technology: A comparison of spectra and color data information

Instrumental test of black tea samples instead of human panel test is attracting massive attention recently. This study focused on an investigation of the feasibility for estimation of the color sensory quality of black tea samples using the VIS-NIR spectroscopy technique, comparing the performances of models based on the spectra and color information. In model calibration, the variables were first selected by genetic algorithm (GA); then the nonlinear back propagation-artificial neural network (BPANN) models were established based on the optimal variables. In comparison with the other models, GA-BPANN models from spectra data information showed the best performance, with the correlation coefficient of 0.8935, and the root mean square error of 0.392 in the prediction set. In addition, models based on the spectra information provided better performance than that based on the color parameters. Therefore, the VIS-NIR spectroscopy technique is a promising tool for rapid and accurate evaluation of the sensory quality of black tea samples.

Intelligent sensing sensory quality of Chinese rice wine using near infrared spectroscopy and nonlinear tools

The approach presented herein reports the application of near infrared (NIR) spectroscopy, in contrast with human sensory panel, as a tool for estimating Chinese rice wine quality; concretely, to achieve the prediction of the overall sensory scores assigned by the trained sensory panel. Back propagation artificial neural network (BPANN) combined with adaptive boosting (AdaBoost) algorithm, namely BP-AdaBoost, as a novel nonlinear algorithm, was proposed in modeling. First, the optimal spectra intervals were selected by synergy interval partial least square (Si-PLS). Then, BP-AdaBoost model based on the optimal spectra intervals was established, called Si-BP-AdaBoost model. These models were optimized by cross validation, and the performance of each final model was evaluated according to correlation coefficient (Rp) and root mean square error of prediction (RMSEP) in prediction set. Si-BP-AdaBoost showed excellent performance in comparison with other models. The best Si-BP-AdaBoost model was achieved with Rp=0.9180 and RMSEP=2.23 in the prediction set. It was concluded that NIR spectroscopy combined with Si-BP-AdaBoost was an appropriate method for the prediction of the sensory quality in Chinese rice wine.