Combining Fourier Transform Mid-Infrared Spectroscopy with Chemometric Methods to Detect Adulterations in Milk Powder.

Adulteration is one of the major concerns among all the quality problems of milk powder. Soybean flour and rice flour are harmless adulterations in the milk powder. In this study, mid-infrared spectroscopy was used to detect the milk powder adulterated with rice flour or soybean flour and simultaneously determine the adulterations content. Partial least squares (PLS), support vector machine (SVM) and extreme learning machine (ELM) were used to establish classification and regression models using full spectra and optimal wavenumbers. ELM models using the optimal wavenumbers selected by principal component analysis (PCA) loadings obtained good results with all the sensitivity and specificity over 90%. Regression models using the full spectra and the optimal wavenumbers selected by successive projections algorithm (SPA) obtained good results, with coefficient of determination (R2) of calibration and prediction all over 0.9 and the predictive residual deviation (RPD) over 3. The classification results of ELM models and the determination results of adulterations content indicated that the mid-infrared spectroscopy was an effective technique to detect the rice flour and soybean flour adulteration in the milk powder. This study would help to apply mid-infrared spectroscopy to the detection of adulterations such as rice flour and soybean flour in real-world conditions.

Identification of Soybean Varieties Using Hyperspectral Imaging Coupled with Convolutional Neural Network.

Soybean variety is connected to stress resistance ability, as well as nutritional and commercial value. Near-infrared hyperspectral imaging was applied to classify three varieties of soybeans (Zhonghuang37, Zhonghuang41, and Zhonghuang55). Pixel-wise spectra were extracted and preprocessed, and average spectra were also obtained. Convolutional neural networks (CNN) using the average spectra and pixel-wise spectra of different numbers of soybeans were built. Pixel-wise CNN models obtained good performance predicting pixel-wise spectra and average spectra. With the increase of soybean numbers, performances were improved, with the classification accuracy of each variety over 90%. Traditionally, the number of samples used for modeling is large. It is time-consuming and requires labor to obtain hyperspectral data from large batches of samples. To explore the possibility of achieving decent identification results with few samples, a majority vote was also applied to the pixel-wise CNN models to identify a single soybean variety. Prediction maps were obtained to present the classification results intuitively. Models using pixel-wise spectra of 60 soybeans showed equivalent performance to those using the average spectra of 810 soybeans, illustrating the possibility of discriminating soybean varieties using few samples by acquiring pixel-wise spectra.

Rapid and Nondestructive Measurement of Rice Seed Vitality of Different Years Using Near-Infrared Hyperspectral Imaging.

Seed vitality is one of the primary determinants of high yield that directly affects the performance of seedling emergence and plant growth. However, seed vitality may be lost during storage because of unfavorable conditions, such as high moisture content and temperatures. It is therefore vital for seed companies as well as farmers to test and determine seed vitality to avoid losses of any kind before sowing. In this study, near-infrared hyperspectral imaging (NIR-HSI) combined with multiple data preprocessing methods and classification models was applied to identify the vitality of rice seeds. A total of 2400 seeds of three different years: 2015, 2016 and 2017, were evaluated. The experimental results show that the NIR-HSI technique has great potential for identifying vitality and vigor of rice seeds. When detecting the seed vitality of the three different years, the extreme learning machine model with Savitzky-Golay preprocessing could achieve a high classification accuracy of 93.67% by spectral data from only eight wavebands (992, 1012, 1119, 1167, 1305, 1402, 1629 and 1649 nm), which could be developed for a fast and cost-effective seed-sorting system for industrial online application. When identifying non-viable seeds from viable seeds of different years, the least squares support vector machine model coupled with raw data and selected wavelengths of 968, 988, 1204, 1301, 1409, 1463, 1629, 1646 and 1659 nm achieved better classification performance (94.38% accuracy), and could be adopted as an optimal combination to identify non-viable seeds from viable seeds.

Near-Infrared Hyperspectral Imaging Combined with Deep Learning to Identify Cotton Seed Varieties.

Cotton seed purity is a critical factor influencing the cotton yield. In this study, near-infrared hyperspectral imaging was used to identify seven varieties of cotton seeds. Score images formed by pixel-wise principal component analysis (PCA) showed that there were differences among different varieties of cotton seeds. Effective wavelengths were selected according to PCA loadings. A self-design convolution neural network (CNN) and a Residual Network (ResNet) were used to establish classification models. Partial least squares discriminant analysis (PLS-DA), logistic regression (LR) and support vector machine (SVM) were used as direct classifiers based on full spectra and effective wavelengths for comparison. Furthermore, PLS-DA, LR and SVM models were used for cotton seeds classification based on deep features extracted by self-design CNN and ResNet models. LR and PLS-DA models using deep features as input performed slightly better than those using full spectra and effective wavelengths directly. Self-design CNN based models performed slightly better than ResNet based models. Classification models using full spectra performed better than those using effective wavelengths, with classification accuracy of calibration, validation and prediction sets all over 80% for most models. The overall results illustrated that near-infrared hyperspectral imaging with deep learning was feasible to identify cotton seed varieties.

Variety Identification of Raisins Using Near-Infrared Hyperspectral Imaging.

Different varieties of raisins have different nutritional properties and vary in commercial value. An identification method of raisin varieties using hyperspectral imaging was explored. Hyperspectral images of two different varieties of raisins (Wuhebai and Xiangfei) at spectral range of 874⁻1734 nm were acquired, and each variety contained three grades. Pixel-wise spectra were extracted and preprocessed by wavelet transform and standard normal variate, and object-wise spectra (sample average spectra) were calculated. Principal component analysis (PCA) and independent component analysis (ICA) of object-wise spectra and pixel-wise spectra were conducted to select effective wavelengths. Pixel-wise PCA scores images indicated differences between two varieties and among different grades. SVM (Support Vector Machine), k-NN (k-nearest Neighbors Algorithm), and RBFNN (Radial Basis Function Neural Network) models were built to discriminate two varieties of raisins. Results indicated that both SVM and RBFNN models based on object-wise spectra using optimal wavelengths selected by PCA could be used for raisin variety identification. The visualization maps verified the effectiveness of using hyperspectral imaging to identify raisin varieties.

Identification of Maize Kernel Vigor under Different Accelerated Aging Times Using Hyperspectral Imaging.

Seed aging during storage is irreversible, and a rapid, accurate detection method for seed vigor detection during seed aging is of great importance for seed companies and farmers. In this study, an artificial accelerated aging treatment was used to simulate the maize kernel aging process, and hyperspectral imaging at the spectral range of 874⁻1734 nm was applied as a rapid and accurate technique to identify seed vigor under different accelerated aging time regimes. Hyperspectral images of two varieties of maize processed with eight different aging duration times (0, 12, 24, 36, 48, 72, 96 and 120 h) were acquired. Principal component analysis (PCA) was used to conduct a qualitative analysis on maize kernels under different accelerated aging time conditions. Second-order derivatization was applied to select characteristic wavelengths. Classification models (support vector machine-SVM) based on full spectra and optimal wavelengths were built. The results showed that misclassification in unprocessed maize kernels was rare, while some misclassification occurred in maize kernels after the short aging times of 12 and 24 h. On the whole, classification accuracies of maize kernels after relatively short aging times (0, 12 and 24 h) were higher, ranging from 61% to 100%. Maize kernels with longer aging time (36, 48, 72, 96, 120 h) had lower classification accuracies. According to the results of confusion matrixes of SVM models, the eight categories of each maize variety could be divided into three groups: Group 1 (0 h), Group 2 (12 and 24 h) and Group 3 (36, 48, 72, 96, 120 h). Maize kernels from different categories within one group were more likely to be misclassified with each other, and maize kernels within different groups had fewer misclassified samples. Germination test was conducted to verify the classification models, the results showed that the significant differences of maize kernel vigor revealed by standard germination tests generally matched with the classification accuracies of the SVM models. Hyperspectral imaging analysis for two varieties of maize kernels showed similar results, indicating the possibility of using hyperspectral imaging technique combined with chemometric methods to evaluate seed vigor and seed aging degree.

Identification of Transgenic Soybean Varieties Using Mid-Infrared Spectroscopy.

Transgenic technology has enormous significance in increasing food production, protecting biodiversity and reducing the use of chemical pesticides and so on. However, there may be some security risks; therefore, research on genetically modified crop identification technology is attracting more and more attention. Mid-infrared spectroscopy combined with feature extraction methods were used to investigate the feasibility of identifying different kinds of transgenic soybeans in the wavelength range of 3 818～734 cm-1. For this purpose, partial least squares-discriminant analysis (PLS-DA) was employed as pattern recognition methods to classify three non-GMO parent soybeans(HC6, JACK and W82)and their transgenic soybeans. The results of the calibration set were 96.67%, 96.67% and 83.33% for three non-GMO parent soybeans and their transgenic soybeans, and the results of the prediction set were 83.33%, 85% and 85%. X-loading weights, variable importance in the projection (VIP) algorithm and second derivative (2-Der) algorithm were applied to select sensitive wavenumbers. The sensitive wavelengths selected with x-loading weights were used to build PLS-DA model, the classification accuracy of the calibration set were 91.11%, 91.67% and 81.67%, and the results of the prediction set were 80%, 80% and 75%. By using the VIP algorithm, the classification accuracy of the calibration set were 94.44%, 95% and 76.67%, and the results of the prediction set were 80%, 85% and 75%. By using the 2-Der algorithm, the classification accuracy of the calibration set were 88.89%, 81.67% and 80%, and the results of the prediction set were 76.67%, 75% and 75%. Principal components analysis (PCA) and independent component analysis (ICA) were applied to extract feature information. The principal components were combined with PLS-DA model. The classification accuracy of the calibration set were 96.67%, 90% and 80%, and the results of the prediction set were 80%, 90% and 80%. The independent components were combined with PLS-DA model. The classification accuracy of the calibration set were 93.33%, 83.33% and 83.33% while the results of the prediction set were 83.33%, 75% and 75%. The overall results indicated that mid-infrared spectroscopy could accurately identify the varieties of the non-GMO parent soybeans, which provided a new idea for nondestructive testing of transgenic soybeans. Feature extraction methods could be used to build more concise models and reduce the amount of program operations combined with sensitive wavenumbers selection methods.

[Fast Detection of Camellia Sinensis Growth Process and Tea Quality Informations with Spectral Technology: A Review].

The research achievements and trends of spectral technology in fast detection of Camellia sinensis growth process information and tea quality information were being reviewed. Spectral technology is a kind of fast, nondestructive, efficient detection technology, which mainly contains infrared spectroscopy, fluorescence spectroscopy, Raman spectroscopy and mass spectroscopy. The rapid detection of Camellia sinensis growth process information and tea quality is helpful to realize the informatization and automation of tea production and ensure the tea quality and safety. This paper provides a review on its applications containing the detection of tea (Camellia sinensis) growing status(nitrogen, chlorophyll, diseases and insect pest), the discrimination of tea varieties, the grade discrimination of tea, the detection of tea internal quality (catechins, total polyphenols, caffeine, amino acid, pesticide residual and so on), the quality evaluation of tea beverage and tea by-product, the machinery of tea quality determination and discrimination. This paper briefly introduces the trends of the technology of the determination of tea growth process information, sensor and industrial application. In conclusion, spectral technology showed high potential to detect Camellia sinensis growth process information, to predict tea internal quality and to classify tea varieties and grades. Suitable chemometrics and preprocessing methods is helpful to improve the performance of the model and get rid of redundancy, which provides the possibility to develop the portable machinery. Future work is to develop the portable machinery and on-line detection system is recommended to improve the further application. The application and research achievement of spectral technology concerning about tea were outlined in this paper for the first time, which contained Camellia sinensis growth, tea production, the quality and safety of tea and by-produce and so on, as well as some problems to be solved and its future applicability in modern tea industrial.

[Fast Identification of Transgenic Soybean Varieties Based Near Infrared Hyperspectral Imaging Technology].

Near-infrared hyperspectral imaging technology combined with chemometrics was applied for rapid and non-invasive transgenic soybeans variety identification. Three different non-GMO parent soybeans(HC6, JACK, TL1)and their transgenic soybeans were chosen as the research object. The developed hyperspectral imaging system was used to acquire the hyperspectral images in the spectral range of 874~1 734 nm with 256 bands of soybeans, and the reflectance spectra were extracted from the region of interest (ROI) in the images. After eliminating the obvious noises, the moving average(MA)was applied as smooth pretreatment, and the wavelengths from 941~1 646 nm were used for later analysis. Partial least squares-discriminant analysis (PLS-DA)was employed as pattern recognition method to class the three different non-GMO parent soybeans. The classification accuracy of both the calibration set and the prediction set were 97.50% and 100% for the HC6, 100% and 100% for the JACK, 96.25% and 92.50% for the TL1, which indicated that hyperspectral imaging technology could identify the varieties of the non-GMO parent soybeans. Then PLS-DA was applied to classify non-GMO parent soybean and its transgenic soybean cultivars for building discriminant models. For the full spectra, the classification accuracy of both the calibration set and the prediction set were 99.17% and 99.17% for the HC6 and its transgenic soybean cultivars, 87.19% and 81.25% for the JACK and its transgenic soybean cultivars, 99.17% and 98.33% for the TL1 and its transgenic soybean cultivars, respectively. The sensitive wavelengths were selected by x-loading weights, and the classification accuracy of the calibration set and prediction set of PLS-DA models based on sensitive wavelengths were 72.50% and 80% for the HC6 and its transgenic soybean cultivars, 80.63% and 79.38% for the JACK and its transgenic soybean cultivars, 85% and 85% for the TL1 and its transgenic soybean cultivars, respectively. These results showed that the pattern recognition for non-GMO parent soybean and their transgenic soybeans was feasible, and the selected sensitive wavelengths could be used for the pattern recognition of non-GMO parent soybeans and transgenic soybeans. The overall results indicated that it was feasible to use near-infrared hyperspectral imaging technology for the pattern recognition of the non-GMO parent soybeans varieties, non-GMO parent soybean and its transgenic soybeans. This study also provided a new alternative for rapid and non-destructive accurate identification of transgenic soybean.

Fruit quality evaluation using spectroscopy technology: a review.

An overview is presented with regard to applications of visible and near infrared (Vis/NIR) spectroscopy, multispectral imaging and hyperspectral imaging techniques for quality attributes measurement and variety discrimination of various fruit species, i.e., apple, orange, kiwifruit, peach, grape, strawberry, grape, jujube, banana, mango and others. Some commonly utilized chemometrics including pretreatment methods, variable selection methods, discriminant methods and calibration methods are briefly introduced. The comprehensive review of applications, which concentrates primarily on Vis/NIR spectroscopy, are arranged according to fruit species. Most of the applications are focused on variety discrimination or the measurement of soluble solids content (SSC), acidity and firmness, but also some measurements involving dry matter, vitamin C, polyphenols and pigments have been reported. The feasibility of different spectral modes, i.e., reflectance, interactance and transmittance, are discussed. Optimal variable selection methods and calibration methods for measuring different attributes of different fruit species are addressed. Special attention is paid to sample preparation and the influence of the environment. Areas where further investigation is needed and problems concerning model robustness and model transfer are identified.

[Rapid detection of nitrogen content and distribution in oilseed rape leaves based on hyperspectral imaging].

Visible and near infrared (Vis-NIR) hyperspectral imaging system was carried out to rapidly determinate the content and estimate the distribution of nitrogen (N) in oilseed rape leaves. Hyperspectral images of 420 leaf samples were acquired at seedling, flowering and pod stages. The spectral data of rape leaves were extracted from the region of interest (ROI) in the wave- length range of 380-1,030 nm. Different spectra preprocessing including Savitzky-Golay smoothing (SG), standard normal variate (SNV), multiplicative scatter correction (MSC), first and second derivatives were applied to improve the signal to noise ratio. Among 471 wavelengths, only twelve wavelengths (467, 557, 665, 686, 706, 752, 874, 879, 886, 900, 978 and 995 nm) were selected by successive projections algorithm(SPA) as the effective wavelengths for N prediction. Based on these effective wavelengths, partial least squares(PLS) and least-squares support vector machines (LS-SVM) calibration models were established for the determination of N content. Reasonable estimation accuracy was obtained, with Rp of 0.807 and RMSEP of 0.387 by PLS and Rp of 0.836 and RMSEP of 0.358 by LS-SVM, respectively. Considering the simple structure and satisfying results of PLS model, SPA-PLS model was used to generate the distribution maps of N content in rape leaves. The concentrations of N were calculated at each pixel of hyperspectral images at the selected effective wavelengths by inputting its correspond- ing spectrum into the established SPA-PLS model. Different colour represented the change in N content in the rape leaves under different fertilizer treatments. By including all pixels within the selected ROI, the average N status can be displayed in more detail and visualised. The visualization of N distribution could be helpful to understanding the change in N content in rape leaves during rape growth period and facilitate discovering the difference of N content within one sample as well as among the samples from different fertilising plots. The overall results revealed that hyperspectral imaging is a promising technique to detect N content and distribution within oilseed rape leaves rapidly and nondestructively.

Potential of visible and near infrared spectroscopy and pattern recognition for rapid quantification of notoginseng powder with adulterants.

Notoginseng is a classical traditional Chinese medical herb, which is of high economic and medical value. Notoginseng powder (NP) could be easily adulterated with Sophora flavescens powder (SFP) or corn flour (CF), because of their similar tastes and appearances and much lower cost for these adulterants. The objective of this study is to quantify the NP content in adulterated NP by using a rapid and non-destructive visible and near infrared (Vis-NIR) spectroscopy method. Three wavelength ranges of visible spectra, short-wave near infrared spectra (SNIR) and long-wave near infrared spectra (LNIR) were separately used to establish the model based on two calibration methods of partial least square regression (PLSR) and least-squares support vector machines (LS-SVM), respectively. Competitive adaptive reweighted sampling (CARS) was conducted to identify the most important wavelengths/variables that had the greatest influence on the adulterant quantification throughout the whole wavelength range. The CARS-PLSR models based on LNIR were determined as the best models for the quantification of NP adulterated with SFP, CF, and their mixtures, in which the rP values were 0.940, 0.939, and 0.867 for the three models respectively. The research demonstrated the potential of the Vis-NIR spectroscopy technique for the rapid and non-destructive quantification of NP containing adulterants.

[Early diagnosis of gray mold on tomato stalks based on hyperspectral data].

Early diagnosis of gray mold on tomato stalks based on hyperspectral data was studied in the present paper. A total of 112 samples' hyperspectral data were collected by hyperspectral imaging system. The study spectral region was from 400 to 1,030 nm. Combined with image processing and chemometric methods, the tomato stalk gray mold diagnosis models were built. Seven effective wavelengths were selected by analysis of variable load distribution in PLS model. The experimental results showed that the excellent results were achieved by EW-LS-SVM model with standard normal variate (SNV) spectral and multiplicative scatter correction (MSC) spectral, and the accuracy of diagnosing gray mold on tomato stalks was satisfied and better than PLS model with whole band. Hence, it is feasible to early diagnose gray mold on tomato stalks using hyperspectral imaging technology, which provides a new early diagnosis and warning method for tomato disease.

Quantitative analysis of total amino acid in barley leaves under herbicide stress using spectroscopic technology and chemometrics.

Visible and near infrared (Vis/NIR) spectroscopy were employed for the fast and nondestructive estimation of the total amino acid (TAA) content in barley (Hordeum vulgare L.) leaves. The calibration set was composed of 50 samples; and the remaining 25 samples were used for the validation set. Seven different spectral preprocessing methods and six different calibration methods (linear and nonlinear) were applied for a comprehensive prediction performance comparison. Successive projections algorithm (SPA) and regression coefficients (RC) were applied to select effective wavelengths (EWs). The results indicated that the latent variables-least-squares-support vector machine (LV-LS-SVM) model achieved the optimal performance. The prediction results by LV-LS-SVM with raw spectra were achieved with a correlation coefficients (r) = 0.937 and root mean squares error of prediction (RMSEP) = 0.530. The overall results showed that the NIR spectroscopy could be used for determination of TAA content in barley leaves with an excellent prediction precision; and the results were also helpful for on-field monitoring of barley growing status under herbicide stress during different growth stages.

Detection of glutamic acid in oilseed rape leaves using near infrared spectroscopy and the least squares-support vector machine.

Amino acids are quite important indices to indicate the growth status of oilseed rape under herbicide stress. Near infrared (NIR) spectroscopy combined with chemometrics was applied for fast determination of glutamic acid in oilseed rape leaves. The optimal spectral preprocessing method was obtained after comparing Savitzky-Golay smoothing, standard normal variate, multiplicative scatter correction, first and second derivatives, detrending and direct orthogonal signal correction. Linear and nonlinear calibration methods were developed, including partial least squares (PLS) and least squares-support vector machine (LS-SVM). The most effective wavelengths (EWs) were determined by the successive projections algorithm (SPA), and these wavelengths were used as the inputs of PLS and LS-SVM model. The best prediction results were achieved by SPA-LS-SVM (Raw) model with correlation coefficient r = 0.9943 and root mean squares error of prediction (RMSEP) = 0.0569 for prediction set. These results indicated that NIR spectroscopy combined with SPA-LS-SVM was feasible for the fast and effective detection of glutamic acid in oilseed rape leaves. The selected EWs could be used to develop spectral sensors, and the important and basic amino acid data were helpful to study the function mechanism of herbicide.

[Analysis of main risk factors causing foodborne diseases in food catering business].

OBJECTIVE: To study main risk factors that cause foodborne diseases in food catering business. METHODS: Data from references and investigations conducted in food catering units were used to establish models which based on @Risk 4.5 with Monte Carlo method referring to food handling practice model (FHPM) to make risk assessment on factors of food contamination in food catering units. The Beta-Poisson models on dose-response relationship to Salmonella (developed by WHO/FAO and United States Department of Agriculture) and Vibrio parahaemolyticus (developed by US FDA) were used in this article to analyze the dose-response relationship of pathogens. RESULTS: The average probability of food poisoning by consuming Salmonella contaminated cooked meat under refrigeration was 1.96 × 10(-4) which was 1/2800 of the food under non-refrigeration (the average probability of food poisoning was 0.35 at room temperature 25°C). The average probability by consuming 6 hours stored meat under room temperature was 0.11 which was 16 times of 2 hours storage (6.79 × 10(-3)). The average probability by consuming contaminated meat without fully cooking was 1.71 × 10(-4) which was 100 times of consuming fully cooked meat (1.88 × 10(-6)). The probability growth of food poisoning by consuming Vibrio parahaemolyticus contaminated fresh seafood was proportional with contamination level and prevalence. CONCLUSION: The primary contamination level, storage temperature and time, cooking process and cross contamination are important factors of catering food safety.

[Detection of pear injury based on visible-near infrared spectroscopy and multispectral image].

A new approach to detect the injury degree and time of pear based on visible-near infrared spectroscopy and multispectral image has been proposed. Firstly, visible-near infrared spectroscopy combined with partial least squares (PLS) and least squares-support vector machine (LS-SVM) was used for pear injury degree and time prediction. The result indicated that these two methods both have good performances in predicting pear injury degree in the late period. The LS-SVM method is more accurate in predicting the injury time of light pear injury, but its overall result of injury time prediction is not as good as that for the PLS method. Then, the multispectral image was used to predict the time of pear injury. The result shows that for different degrees of pear injury, the prediction models based on LS-SVM have a better performance with correlation coefficients around 5.85. The result of this study can be used to detect the injury degree and time of pear rapidly and non-destructively, and provide a new approach to pear classification.

[Quickly determination of titanium dioxide content in juice based on Vis/NIR spectroscopy technique].

In order to quickly and accurately detect the content of titanium dioxide in the juice, a method combining chemometrics and Vis/NIR spectroscopy technique was used in the present study. First, the content of titanium dioxide in the juice sample was determined by using spectrophotometer and standard curve of titanium dioxide. Then, different amount of pure titanium dioxide was adulterated into the juice collected from the market to prepare eight different content samples. A total of 320 juice samples were studied. Two hundred samples (25 samples for each content) were randomly selected from the 320 samples to be the calibration set while the other 120 samples (15 samples for each content) were selected as the validation set. The spectra of juice were within near infrared (NIR) and mid-infrared (MIR). First six different preprocessing methods were compared, such as standard normal variate (SNV), moving average, derivative and multivariate scatter correction (MSC). The optimal partial least squares(PLS)was built after the performance comparison of different preprocessing methods. Another algorithm, principal component-artificial neural network (PC-ANN), was also used: first, the original spectral date was processed using principal component analysis, the best number of principal components (PCs) was selected, and the scores of these PCs would be taken as the input of the artificial neural network (ANN). The PC-ANN was trained with samples in the calibration collection and the samples in prediction set were predicted. After comparison, MSC was found to be the most appropriate spectral preprocessing method and the best number of PCs is 7. The correlation coefficients (R2) between the real values and predicted ones by discriminant analysis model were 0.9008 (PLS) and 0.8684 (PC-ANN) respectively. The root mean standard errors of prediction (RMSEP) by PLS and PC-ANN were 0.05 (PLS) and 0.04 (PC-ANN) respectively. The result indicated that the content of titanium dioxide in the juice powder to be quickly detected by nondestructive determination method was very feasible and laid a solid foundation for setting up the titanium dioxide content forecasting model of juice powder.

Application of near-infrared hyperspectral imaging to identify a variety of silage maize seeds and common maize seeds.

Common maize seeds and silage maize seeds are similar in appearance and are difficult to identify with the naked eye. Four varieties of common maize seeds and four varieties of silage maize seeds were identified by near-infrared hyperspectral imaging (NIR-HSI) combined with chemometrics. The pixel-wise principal component analysis was used to distinguish the differences among different varieties of maize seeds. The object-wise spectra of each single seed sample were extracted to build classification models. Support vector machine (SVM) and radial basis function neural network (RBFNN) classification models were established using two different classification strategies. First, the maize seeds were directly classified into eight varieties with the prediction accuracy of the SVM model and RBFNN model over 86%. Second, the seeds of silage maize and common maize were firstly classified with the classification accuracy over 88%, then the seeds were classified into four varieties, respectively. The classification accuracy of silage maize seeds was over 98%, and the classification accuracy of common maize seeds was over 97%. The results showed that the varieties of common maize seeds and silage maize seeds could be classified by NIR-HSI combined with chemometrics, which provided an effective means to ensure the purity of maize seeds, especially to isolate common seeds and silage seeds.

Discrimination of Grape Seeds Using Laser-Induced Breakdown Spectroscopy in Combination with Region Selection and Supervised Classification Methods.

The wine-making industry generates a considerable amount of grape pomace. Grape seeds, as an important part of pomace, are rich in bioactive compounds and can be reutilized to produce useful derivatives. The nutritional properties of grape seeds are largely influenced by the cultivar, which calls for effective identification. In the present work, the spectral profiles of grape seeds belonging to three different cultivars were collected by laser-induced breakdown spectroscopy (LIBS). Three conventional supervised classification methods and a deep learning method, a one-dimensional convolutional neural network (CNN), were applied to establish discriminant models to explore the relationship between spectral responses and cultivar information. Interval partial least squares (iPLS) algorithm was successfully used to extract the spectral region (402.74-426.87 nm) relevant for elemental composition in grape seeds. By comparing the discriminant models based on the full spectra and the selected spectral regions, the CNN model based on the full spectra achieved the optimal overall performance, with classification accuracy of 100% and 96.7% for the calibration and prediction sets, respectively. This work demonstrated the reliability of LIBS as a rapid and accurate approach for identifying grape seeds and will assist in the utilization of certain genotypes with desirable nutritional properties essential for production rather than their being discarded as waste.