Residual Predictive Deviation Values Research Articles

Detection of Anthocyanins in Potatoes Using Micro-Hyperspectral Images Based on Convolutional Neural Networks.

The color potato has the function of both a food and vegetable. The color potato not only contains various amino acids and trace elements needed by the human body but also contains anthocyanins. Anthocyanins have many functions, such as antioxidation, inflammation inhibition, vision improvement, and cancer prevention, so colored potatoes are deeply loved by consumers and have good market prospects. However, at present, the detection of anthocyanin content in color potatoes mainly depends on chemical methods, which are time-consuming and laborious, so it is necessary to study a fast and accurate detection method. In this study, microscopic hyperspectral equipment was used to collect the spectral information of the outer skin and inner skin of potatoes. The original spectrum, pretreatment spectrum, and characteristic spectrum variables of the outer skin and inner skin were predicted by the convolution neural network (CNN) algorithm and partial least squares regression (PLS) algorithm, respectively, and the performance of the model was evaluated by the prediction set correlation coefficient (Rp), prediction set root mean square error (RMSEP), correction set correlation coefficient (Rc), correction set root mean square error (RMSEC), and residual prediction deviation (RPD). The results revealed that the inner skin Raw + CNN model constructed under raw spectral data is optimal with Rc = 0.9508, RMSEC = 0.0374%, Rp = 0.9461, RMSEP = 0.2361% and RPD = 4.4933. The inner skin Savitzky-Golay (SG) + Detrend (DET) + CNN model constructed from pre-processed spectral data is optimal with Rc = 0.9499, RMSEC = 0.0359%, Rp = 0.9439, RMSEP = 0.2384%, RPD = 4.6516. The inner skin DET + competitive adaptive reweighted sampling (CARS) +CNN model constructed from the feature-based spectral data was optimal with Rc = 0.9527, RMSEC = 0.0708%, Rp = 0.9457, RMSEP = 0.2711%, and RPD = 4.1623. It can be seen that the Rp, RMSEP, Rc, RMSEC, and RPD values for modeling the spectral information of the inner skin were higher than those of the outer skin under the three different spectral data. The prediction accuracy of the model built by the CNN algorithm was better than the conventional algorithm PLS, the application of the CNN algorithm in inner skin can achieve accurate prediction of anthocyanin content in potato.

Development of machine learning models using multi-source data for geographical traceability and content prediction of Eucommia ulmoides leaves

Rapid and scientific quality evaluation is a hot topic in the research of food and medicinal plants. With the increasing popularity of derivative products from Eucommia ulmoides leaves, quality and safety have attracted public attention. The present study utilized multi-source data and traditional machine learning to conduct geographical traceability and content prediction research on Eucommia ulmoides leaves. Explored the impact of different preprocessing methods and low-level data fusion strategy on the performance of classification and regression models. The classification analysis results indicated that the partial least squares discriminant analysis (PLS-DA) established by low-level fusion of two infrared spectroscopy techniques based on first derivative (FD) preprocessing was most suitable for geographical traceability of Eucommia ulmoides leaves, with an accuracy rate of up to 100 %. Through regression analysis, it was found that the preprocessing methods and data blocks applicable to the four chemical components were inconsistent. The optimal partial least squares regression (PLSR) model based on aucubin (AU), geniposidic acid (GPA), and chlorogenic acid (CA) had a residual predictive deviation (RPD) value higher than 2.0, achieving satisfactory predictive performance. However, the PLSR model based on quercetin (QU) had poor performance (RPD = 1.541) and needed further improvement. Overall, the present study proposed a strategy that can effectively evaluate the quality of Eucommia ulmoides leaves, while also providing new ideas for the quality evaluation of food and medicinal plants.

Comparative study of Vis/NIR reflectance and transmittance method for on-line detection of strawberry SSC

Strawberry, as a fragile and vulnerable fruit, the realization of automatic sorting is conducive to improve the intelligent level of strawberry industry and improve the ability of product quality management. An on-line soluble solids content (SSC) detection prototype which can protect the strawberry from mechanical damage was researched and developed. The reflectance and transmittance of visible and near infrared (Vis/NIR) spectra were acquired by the prototype respectively, and the performances of the two spectra on the SSC detection performance of strawberry were compared. Four feature selection algorithms like competitive adaptive reweighted sampling (CARS) ware used for reflectance and transmittance spectra to reduce the spectra complexity, improve the strawberry SSC detection accuracy and optimize the running time of the prototype. The comparison showed that the transmittance spectra can reflect the internal SSC information of strawberry better. Then the results of feature variable selection showed that strawberry transmittance spectra combined with CARS algorithm achieved the best result of SSC prediction, and the prediction correlation coefficient (Rp) was 0.928, the root mean square error of prediction (RMSEP) was 0.412 °Brix, and the residual predictive deviation (RPD) value was 2.670. The CARS-PLS model for reflectance spectra also obtained the optimization result in the reflectance group, but its Rp, RMSEP and RPD value was 0.812, 0.587 °Brix and 1.670 respectively, which still did not meet the reliability of application. The results demonstrated that the Vis/NIR transmittance spectra have great application potential in strawberry on-line internal quality detection.

Application of near-infrared spectroscopy and chemometrics for the rapid detection of insect protein adulteration from a simulated matrix

The popularity of insect protein as a food and feed supplement is growing. Protein quality, end use and prices vary considerably between different insect species, which may incentivise insect protein adulteration. Here, near-infrared (NIR) spectroscopy and chemometrics were used to detect the presence of cricket, black soldier fly larvae (BSFL) and mealworm proteins in a simulated complex insect protein mixture. Additionally, BSFL protein powders collected from three commercial sources were investigated to determine whether the NIR-based technology can discriminate the proteins obtained from different companies based on their composition. The proximate analysis suggests compositional protein, fat and chitin differences between insect species. A partial least square (PLS) regression model obtained Q2 values ranging from 0.991 to 0.997 for the predictions of the content of protein mixtures containing BSFL, cricket and mealworm powders mixed at various proportions. The root mean square error of cross-validation (RMSCV) values range from 1.8% to 2.9%, and residual prediction deviation (RPD) values from 10.4 to 17.1 for the adulterated insect protein powders. The accuracy of the prediction model (∼2%) for the adulterated percentages varied depending on the insect species. The NIR spectra could differentiate (Q2 = 0.999) the origin of BSFL protein powders from three different companies and two types of processing (whole meal and defatted samples). Overall, this study established a rapid and low-cost insect protein adulteration monitoring pipeline for the three common insect protein powders. We envisage that NIR can be applied to assess insect adulteration, authentication, and quality control in the emerging insect food and feed industries.

Detection of saffron adulteration with Crocus sativus style using NIR-hyperspectral imaging and chemometrics

Saffron is a valuable spice that is often adulterated. This study proposes using near-infrared hyperspectral imaging (NIR-HSI) and chemometrics as a fast and cost-effective method for detecting and quantifying adulteration in saffron stigmas. Adulterated saffron samples were prepared by adding Crocus sativus style to pure saffron stigmas in varying concentrations (20–90%). The spectral data were pre-treated using standard normal variate (SNV), and multiplicative scatter correction (MSC), while variable reduction was performed by Principal Component Analysis (PCA) and Partial Least Squares (PLS). Classification was done using Linear Discriminant Analysis (LDA), PLS-DA, Support Vector Machine (SVM), and Multi-layer Perceptron (MLP) models, while quantification was achieved by PLS, PCA, SVM, and MLP-based regression models. The HSI technique achieved correct classification rates of 95.6%–100% in discriminating authentic saffron from plant adulterants and adulterated saffron across all the models. Regression models to quantify the percentage style adulteration in saffron demonstrated excellent prediction abilities with almost all models achieving RPD (Residual Predictive Deviation) values of 3.0–5.4. The MLP model (1 hidden layer with 3 neurons) built from SNV pre-processed and PLS reduced data (15 LVs), showed exceptional predictive capabilities, with an R2p of 0.97, a Root Mean Squared Error of Prediction (RMSEP) of 4.3%, and an RPD of 5.4. The results demonstrate the potential of NIR-HSI and chemometrics for rapid and nondestructive detection and quantification of style in saffron stigmas.

Flexible Vis/NIR sensing system for banana chilling injury

The transportation stage of the banana production chain has influence on banana quality and other lifecycle parameters. Certain factors, such as chilling injury (CI), have detrimental effects on bananas, leading to increased losses. The commonly used monitoring and evaluation methods, such as machine vision, face challenges in achieving real-time detection and are susceptible to environmental interference, leading to deviations in the results. Therefore, a flexible visible (Vis)/near-infrared (NIR) real-time sensing system (FVN) was developed for real-time monitoring of the CI status of bananas. The color space of bananas was analyzed and predicted based on the multiple linear regression (MLR) model, and the results showed that the coefficient of determination (R2p) of a* of 0.97, and the residual prediction deviation (RPD) of 4.95, which indicated that the prediction of a* reached a high accuracy. The RPD values for the predictions of L* and b* exceeding 2.5 indicate that the FVN based on the MLR model could be applicable to the majority of market demands. The self-developed classification prediction model (SCP) exhibits evident advantages in predicting the occurrence and elapsed duration of CI, with prediction accuracies of 98.3 % and 95.5 %. In addition, a comprehensive comparative analysis of the FVN is carried out in terms of power consumption and cost, highlighting its great advantages. The application of FVN can effectively reduce the waste of bananas in the market supply chain, greatly alleviate the problem of unpredictable fruit chilling damage, and thus promote more sustainable and cleaner production in the banana industry.

Flexible wireless in situ optical sensing system for banana ripening monitoring

AbstractDetermining the ripening level of bananas in the environment of banana ripening plays an important role in ensuring the quality of bananas, reducing losses and waste, and improving economic benefits. Traditional monitoring methods face a series of challenges such as inefficiency, high energy consumption, and high cost. This paper aims to develop a flexible wireless in situ optical sensing system (FIS) for banana ripening monitoring with a good prediction effect, low energy consumption, and high economy. When monitoring the ripening process, the ripeness of the bananas can be accurately determined without taking out the bananas. Among them, the prediction effect on soluble solid content and a* is the best, the residual predictive deviation value is above 3, and the classification accuracy is as high as 95%. Compared to other commercially available spectrometers, the device exhibits the lowest power consumption characteristics. The theoretical maximum energy consumption of the FIS is only 627 mW, and a single measurement only consumes 15 J. The overall price of the FIS does not exceed $60. The application of the FIS can effectively reduce the waste of bananas during the ripening process, and greatly alleviate the labor‐intensive and inefficient problems of fruit maturity monitoring, thereby promoting more sustainable and cleaner production in the banana industry.

Rapid and comprehensive grade evaluation of Keemun black tea using efficient multidimensional data fusion

To develop a comprehensive evaluation method for Keemun black tea, we used micro-near-infrared spectroscopy, computer vision, and colorimetric sensor array to collect data. We used support vector machine, least-squares support vector machine (LS-SVM), extreme learning machine, and partial least squares discriminant analysis algorithms to qualitatively discriminate between different grades of tea. Our results indicated that the LS-SVM model with mid-level data fusion attained an accuracy of 98.57% in the testing set. To quantitatively determine flavour substances in black tea, we used support vector regression. The correlation coefficient for the predicted sets of gallic acid, caffeine, epigallocatechin, catechin, epigallocatechin gallate, epicatechin, gallocatechin gallate and total catechins were 0.84089, 0.94249, 0.94050, 0.83820, 0.81111, 0.82670, 0.93230, and 0.93608, respectively. Furthermore, all compounds exhibited residual predictive deviation values exceeding 2. Hence, combining spectral, shape, colour, and aroma data with mid-level data can provide a rapid and comprehensive assessment of Keemun black tea quality.

Spatio-temporal distribution patterns and quantitative detection of aflatoxin B1 and total aflatoxin in peanut kernels explored by short-wave infrared hyperspectral imaging.

Aflatoxin contamination in peanut kernels seriously harms the health of humans and causes significant economic losses. Rapid and accurate detection of aflatoxin is necessary to minimize its contamination. However, current detection methods are time-consuming, expensive and destructive to samples. Therefore, short-wave infrared (SWIR) hyperspectral imaging coupled with multivariate statistical analysis was used to investigate the spatio-temporal distribution patterns of aflatoxin, and quantitatively detect the aflatoxin B1 (AFB1) and total aflatoxin in peanut kernels. In addition, Aspergillus flavus contamination was identified to prevent the production of aflatoxin. The result of validation set demonstrated that SWIR hyperspectral imaging could predict the contents of the AFB1 and total aflatoxin accurately, with residual prediction deviation values of 2.7959 and 2.7274, and limits of detection of 29.3722 and 45.7429μg/kg, respectively. This study presents a novel method for the quantitative detection of aflatoxin and offers an early warning system for its potential application.

Determination of glucosinolates in broccoli (Brassica oleracea var. italica) by combining mid‐infrared (MIR) spectroscopy with chemometrics

SummaryGlucosinolates (GSLs), compounds with potential chemo‐preventative properties, can be affected by pre‐ and post‐harvest processes. Therefore, monitoring their content is important for the food and agricultural industries. This study evaluates the application of mid‐infrared (MIR) spectroscopy for the determination of GSLs in commercial broccoli (Brassica oleracea var. italica). Broccoli (n = 53) from different varieties were analysed using MIR spectroscopy and high‐performance liquid chromatography to develop partial least squares models for six individual GLSs, total indolic glucosinolates and total GSLs. The coefficient of determination in cross‐validation () of 0.50–0.78 whereas residual predictive deviation (RPD) values of 1.35–2.19 for different GSLs. Cross‐validation models developed using Tenderstem® broccoli yielded an of 0.41–0.91 and an RPD of 0.81–2.97. This study showed that MIR spectroscopy can be considered for the determination of some broccoli GSLs. Differences in the loadings between the models demonstrated that variations in broccoli composition can influence the cross‐validation models for GSLs.

Rheo-fermentation properties of bread dough with different gluten contents processed by 3D printing

The rheological properties of dough closely correlate to a dough’s ability to be three-dimensionally (3D) printed, but only weakly characterize its fermentation and baking process. This study aimed to use rheo-fermentation properties to predict rheological properties of dough, thereby obtaining indirect information on both 3D printing properties and post-processing characteristics. The 3D printing behavior and baking quality of the dough were measured. A gluten content of 13% was found to be the most suitable for 3D printing and exhibited desirable performance during fermentation and baking. Pearson correlation analysis revealed a strong correlation between rheological properties and rheo-fermentation properties. Using partial least squares regression-based models, the coefficients of determination of the prediction for rheological parameters (G′, G″, η*) were 0.920, 0.854 and 0.863, respectively, with corresponding residual prediction deviation values of 3.063, 3.774, and 4.773. These findings suggest that 3D printing of bread dough products can be easily and successfully accomplished.

Sorting of navel orange soluble solids content based on online near infrared spectroscopy

Abstract A rapid, nondestructive, and online detection of the internal quality of navel orange cannot only reduce the labor intensity, but also improve the economic benefits of the navel orange. In this paper, an online detection and sorting equipment is designed for navel orange. The transmission spectrum data of 1697 navel oranges are divided into the calibration, prediction, and validation sets, with a ratio of 14:3:3. Pre-processing methods such as first derivative (FD), second derivative (SD), standard normal variate transform (SNV), and multiplicative scatter correction (MSC) were chosen to process the spectra. Accordingly, the soluble solids content prediction model for navel oranges is established using standard normal variable transformation (SNV) and partial least squares (PLS). The determination coefficients (R 2) of the calibration set, prediction set, and validation set are 0.8476, 0.8326, and 0.8025, respectively. Moreover, the corresponding root mean square errors are 0.5097°Brix, 0.5590°Brix, and 0.6048°Brix, respectively. The residual predictive deviation (RPD) value is 2.4510 (i.e., greater than 2.0), indicating that the model performs accurate predictive simulations, and has high reliability. In addition, two classification methods based on the national standard method and the normal probability graph of the soluble solids content of navel oranges were used to classify navel oranges into three classes for online validation. 185 navel oranges were selected for online validation, in which the classification method based on the normal probability graph of the soluble solids content of navel oranges was more effective and its average sorting accuracy was 81.13 %. Likewise, the mean absolute error (MAE) is 0.4613°Brix. The experimental results show that the online sorting equipment possesses high sorting accuracy and can be practically used for actual postharvest processing.

Comparison of Si-GA-PLS and Si-CARS-PLS build algorithms for quantitation of total polyphenols in black tea using the spectral analytical system.

The objective of the current study was to compare two machine learning approaches for the quantification of total polyphenols by choosing the optimal spectral intervals utilizing the synergy interval partial least squares (Si-PLS) model. To increase the resilience of built models, the genetic algorithm (GA) and competitive adaptive reweighted sampling (CARS) were applied to a subset of variables. The collected spectral data were divided into 19 sub-interval selections totaling 246 variables, yielding the lowest root mean square error of cross-validation (RMSECV). The performance of the model was evaluated using the correlation coefficient for calibration (RC ), prediction (RP ), RMSECV, root mean square error of prediction (RMSEP) and residual predictive deviation (RPD) value. The Si-GA-PLS model produced the following results: PCs = 9; RC = 0.915; RMSECV = 1.39; RP = 0.8878; RMSEP = 1.62; and RPD = 2.32. The performance of the Si-CARS-PLS model was noted to be best at PCs = 10, while RC = 0.9723, RMSECV = 0.81, RP = 0.9114, RMSEP = 1.45 and RPD = 2.59. The build model's prediction ability was amended in the order PLS < Si-PLS < CARS-PLS when full spectroscopic data were used and Si-PLS < Si-GA-PLS < Si-CARS-PLS when interval selection was performed with the Si-PLS model. Finally, the developed method was successfully used to quantify total polyphenols in tea. © 2023 Society of Chemical Industry.

Evaluation of the optical layout and sample size on online detection of apple watercore and SSC using Vis/NIR spectroscopy

Visible near-infrared (Vis/NIR) spectroscopy has the capability of identifying apple watercore disorder online. However, the various distributions of watercore tissues had significant effects on the prediction of soluble solids content (SSC) and watercore severity index (WSI). In this study, based on the optical simulation method, two optical configurations were evaluated and applied to two batch samples. Under the dual-light source layout, the partial least squares regression (PLSR) model demonstrated superior prediction performance for SSC of identical-sized apples (batch 1). The root mean square error of the calibration set (RMSEC) was 0.34%, the root mean square error of the prediction set (RMSEP) was 0.37%, and the residual predictive deviation (RPD) was 3.78. RMSEC, RMSEP, and RPD for the PLSR-based WSI prediction model were 2.00%, 1.82%, and 1.69. While for apples of varying sizes (batch 2), with better RMSEC, RMSEP and RPD values of 0.35%, 0.43%, and 3.58 for SSC and 5.01%, 5.00% and 1.21 for WSI than in dual-light source layout, the four-light source layout outperformed the dual-light source layout. The experimental findings demonstrated that a detection system with a four-light source layout can mitigate the effects of apple size and watercore distribution on the efficacy of SSC and WSI prediction.

Evaluation of mid-infrared spectra associated with chemometrics for the determination of physicochemical properties during fermentation of a new strawberry-based beverage with water kefir grains

This study aimed to explore the mid-infrared technique for the characterization of a non-dairy fermented beverage. Fruit juice was used as a substrate for fermentation made by microorganisms from water kefir. The fermented beverage was evaluated by standardized methods regarding pH, total titratable acidity, soluble solids, and total reducing sugars. Then, infrared spectra were obtained and calibration models were built to predict these parameters in the fermented beverage. All obtained models showed high coefficients of determination values (R2) in the calibration and prediction stages, in addition to high residual predictive deviation (RPD) values. For the prediction data, the values of R² were equal to 0.91, 0.93, 0.95, and 0.91 for pH, total titratable acidity, soluble solids, and total reducing sugars, respectively. It was evidenced that the soluble solids content was the most suitable physicochemical property for estimation by the elaborated calibration models (higher R2 and RPD values). The results showed that such parameters could be estimated in kefir quickly through the mid-infrared analysis.

Predicting Chemical Composition and Apparent Total Tract Digestibility on Freeze-Dried Not Ground Faeces Using Near-Infrared Spectroscopy in Pigs

The present study aimed to compare NIRS results using freeze-dried ground or not ground (FDG or FDNG) faeces to predict faecal chemical composition and apparent total tract digestibility (ATTD) coefficients. Two different batches of pigs were used (n = 20 mixed sex pens/batch; 11 pigs/pen; Duroc × (Large White × Landrace)). The first batch of pigs (B1; 50.1 ± 3.44 kg body weight (BW)) was used at 13 wks of age and the second batch (B2; 87.0 ± 4.10 kg BW) was used at 18 wks of age. For both B1 and B2, pens were assigned to five diets formulated to obtain a control [10.03 MJ of net energy (NE), 160.0 g of crude protein (CP), and 9.5 g of standardized ileal digestive (SID) lysine (Lys) per kg of feed], low protein (132.0 g CP and 7.5 g SID Lys), high protein (188.0 g CP and 11.5 g SID Lys), low energy (9.61 MJ NE/kg), and high energy (10.45 MJ NE/kg) diets. After a 10-day adaptation period, one faecal sample was collected daily from each pen floor during 6 days in both B1 and B2 (n = 120/batch). Faecal samples were freeze-dried and analysed via NIRS as FDNG and FDG faeces. Dry matter (DM), organic matter (OM), CP, gross energy (GE), fat, and ATTD coefficients were analysed/calculated. The NIRS calibrations were evaluated by cross-validation, splitting the data in four random groups, or using the leave-one-out method. For both FDNG and FDG faeces, coefficients of determination for calibration (R2cv) and residual predictive deviation (RPD) values were: close to 0.9 and 3 for DM and CP, 0.7–0.8 and ≥2 for OM and GE, 0.6 and &lt;2 for fat, and 0.54–0.75 and ≤2 for ATTD coefficients, respectively. CP was better predicted using FDG faeces (p &lt; 0.05), while DM and OM ATTD were better predicted using FDNG faeces (p &lt; 0.05). In conclusion, NIRS successfully predicts faeces’ chemical components and ATTD coefficients of nutrients using FDNG or FDG faeces.

Potential of Near-Infrared Spectroscopy (NIRS) for Efficient Classification Based on Postharvest Storage Time, Cultivar and Maturity in Coconut Water.

Coconut water (CW) is a popular and healthful beverage, and ensuring its quality is crucial for consumer satisfaction. This study aimed to explore the potential of near-infrared spectroscopy (NIRS) and chemometric methods for analyzing CW quality and distinguishing samples based on postharvest storage time, cultivar, and maturity. CW from nuts of Wenye No. 2 and Wenye No. 4 cultivars in China, with varying postharvest storage time and maturities, were subjected to NIRS analysis. Partial least squares regression (PLSR) models were developed to predict reducing sugar and soluble sugar contents, revealing moderate applicability but lacking accuracy, with the residual prediction deviation (RPD) values ranging from 1.54 to 1.83. Models for TSS, pH, and TSS/pH exhibited poor performance with RPD values below 1.4, indicating limited predictability. However, the study achieved a total correct classification rate exceeding 95% through orthogonal partial least squares discriminant analysis (OPLS-DA) models, effectively discriminating CW samples based on postharvest storage time, cultivar, and maturity. These findings highlight the potential of NIRS combined with appropriate chemometric methods as a valuable tool for analyzing CW quality and efficiently distinguishing samples. NIRS and chemometric techniques enhance quality control in coconut water, ensuring consumer satisfaction and product integrity.

Prediction of quality traits in dry pepper powder using visible and near-infrared spectroscopy

Fruit quality phenotyping is a bottleneck in plant breeding. The present work aimed to investigate the applicability of visible (Vis) and near-infrared (NIR) spectroscopy for quality evaluation in dry red chili powder. We constructed prediction models for the American Spice Trade Association (ASTA)-colour and the Scoville Heat Unit (SHU)-pungency pepper traits using spectroscopy and multivariate statistical techniques. Predictive partial least squares (PLS) models were successfully achieved with high correlations (r) between the predicted and reference values for calibration and validation (r = 0.955 and 0.928 for ASTA-colour; r = 0.941 and 0.918 for SHU-pungency). Spectroscopy data from visible and short-wave radiation (Vis-SWNIR) provided the most robust (residual predictive deviation value) model for ASTA-colour (RPD = 2.84) and long-wave radiation (LWNIR) for SHU-pungency (RPD = 2.48). Spectral categories for wavelength range selection, variable importance for effective wavelength selection, and root mean press-statistic for factor selection were important criteria for PLS. Trait variance and distribution were also important criteria for the predictive capacity and power of the models. In conclusion, non-invasive spectroscopy was a promising tool in our study for dry red chili quality phenotyping.

Quantitative analysis of index factors in agricultural compost by infrared spectroscopy

Hyperspectral technology, with its high spectrum resolution and nanometer continuous spectral information acquisition ability, provide a possibility for rapidly and nondestructive evaluating compost maturity. In this study, the near-infrared spectroscopy (NIRS) analysis techniques was used to analyze quantitatively organic matter (OM) content, total nitrogen (TN) content and carbon-nitrogen (C/N) ratio in compost based on two different composting procedures. In the basis of spectra preprocessing and strategies of variable selection, the nonlinear modeling LBC-siPLS-PLSR for OM, MSC-SPA-PLSR for TN and R-SPA-PLSR for C/N ratio was respectively constructed using partial least squares regression (PLSR). LBC-siPLS-PLSR, MSC-SPA-PLSR and R-SPA-PLSR provided a better prediction capability with root mean square error of prediction, the coefficient of determination for prediction and residual predictive deviation values of 4.061, 0.746 and 2.02 for OM, values of 0.205, 0.65 and 1.71 for TN and values of 1.11, 0.706 and 2.07 for C/N ratio, respectively. These results showed that the NIRS technique could be fitted to each element, using specific spectrum pretreatment, in order to achieve an acceptable accuracy in the prediction.

Rapid and Nondestructive Identification of Origin and Index Component Contents of Tiegun Yam Based on Hyperspectral Imaging and Chemometric Method

Tiegun yam is a typical food and medicine agricultural product, which has the effects of nourishing the kidney and benefitting the lungs. The quality and price of Tiegun yam are affected by its origin, and counterfeiting and adulteration are common. Therefore, it is necessary to establish a method to identify the origin and index component contents of Tiegun yam. Hyperspectral imaging combined with chemometrics was used, for the first time, to explore and implement the identification of origin and index component contents of Tiegun yam. The origin identification models were established by partial least squares-discriminant analysis (PLS-DA), support vector machine (SVM), and random forest (RF) using full wavelength and feature wavelength. Compared with other models, MSC-PLS-DA is the best model, and the accuracy of the training set and prediction set is 100% and 98.40%. Partial least squares regression (PLSR), random forest (RF), and support vector regression (SVR) models were used to predict the contents of starch, polysaccharide, and protein in Tiegun yam powder. The optimal residual predictive deviation (RPD) values of starch, polysaccharide, and protein prediction models selected in this study were 5.21, 3.21, and 2.94, respectively. The characteristic wavelength extracted by the successive projections algorithm (SPA) method can achieve similar results as the full-wavelength model. These results confirmed the application of hyperspectral imaging (HSI) in the identification of the origin and the rapid nondestructive prediction of starch, polysaccharide, and protein contents of Tiegun yam powder. Therefore, the HSI combined with the chemometric method was available for conveniently and accurately determining the origin and index component contents of Tiegun yam, which can expect to be an attractive alternative method for identifying the origin of other food.