Correlation Coefficient Of Prediction Set Research Articles

Spoilage Monitoring and Early Warning for Apples in Storage Using Gas Sensors and Chemometrics.

In the process of storage and cold chain logistics, apples are prone to physical bumps or microbial infection, which easily leads to spoilage in the micro-environment, resulting in widespread infection and serious post-harvest economic losses. Thus, development of methods for monitoring apple spoilage and providing early warning of spoilage has become the focus for post-harvest loss reduction. Thus, in this study, a spoilage monitoring and early warning system was developed by measuring volatile component production during apple spoilage combined with chemometric analysis. An apple spoilage monitoring prototype was designed to include a gas monitoring array capable of measuring volatile organic compounds, such as CO2, O2 and C2H4, integrated with the temperature and humidity sensor. The sensor information from a simulated apple warehouse was obtained by the prototype, and a multi-factor fusion early warning model of apple spoilage was established based on various modeling methods. Simulated annealing-partial least squares (SA-PLS) was the optimal model with the correlation coefficient of prediction set (Rp) and root mean square error of prediction (RMSEP) of 0.936 and 0.828, respectively. The real-time evaluation of the spoilage was successfully obtained by loading an optimal monitoring and warning model into the microcontroller. An apple remote monitoring and early warning platform was built to visualize the apple warehouse's sensors data and spoilage level. The results demonstrated that the prototype based on characteristic gas sensor array could effectively monitor and warn apple spoilage.

Near infrared spectroscopy for simultaneous quantification of five chemical components in Arnebiae Radix (AR) with partial least squares and support vector machine algorithms

Arnebiae Radix (AR) has high economic and medicinal value, and it is used as Chinese herbal medicine in clinic for a long time. Nowadays, the quality assessment of AR is generally performed on expensive instruments with time-consuming and laborious methods. In this study, the near-infrared (NIR) spectroscopy technique was applied to develop a rapid, simple and accurate method for quantifying five main chemical components including acetylshikonin, β-acetoxyisovalerylalkannin, isobutylshikonin, isovalerylshikonin and β,β-dimethylacrylalkannin in AR coupled with chemometrics, aiming at rapidly evaluating the quality of AR. NIR spectra of 110 batches of AR samples were collected to quantify the content of five chemical components. The partial least squares (PLS) and support vector machine (SVM) algorithms were separately used to develop the calibration models. Different spectral pretreatment methods and feature selection methods were compared in details during the model development. As a result, the performances of PLS models for quantifying the five chemical components were superior to SVM models. The correlation coefficients of calibration set (rcal) of PLS models for acetylshikonin, β-acetoxyisovalerylalkannin, isobutylshikonin, isovalerylshikonin and β,β-dimethylacrylalkannin were 0.9972, 0.9973, 0.9927, 0.9981 and 0.9977, respectively, while the correlation coefficients of prediction set (rpre) were 0.9913, 0.9855, 0.9825, 0.9916 and 0.9825, respectively. Besides, the relative prediction deviation (RPD) of PLS models reached 7.0, 5.85, 4.92, 7.52 and 5.43, respectively. This study demonstrated that the NIR spectroscopy technique was a promising approach as routine analysis of AR, and it could be an alternative method for rapid quality assessment of AR.

Robustness of calibration model for prediction of lignin content in different batches of snow pears based on NIR spectroscopy.

Snow pear is very popular in southwest China thanks to its fruit texture and potential medicinal value. Lignin content (LC) plays a direct and negative role (higher concentration and larger size of stone cells lead to thicker pulp and deterioration of the taste) in determining the fruit texture of snow pears as well as consumer purchasing decisions of fresh pears. In this study, we assessed the robustness of a calibration model for predicting LC in different batches of snow pears using a portable near-infrared (NIR) spectrometer, with the range of 1033-2300 nm. The average NIR spectra at nine different measurement positions of snow pear samples purchased at four different periods (batch A, B, C and D) were collected. We developed a standard normal variate transformation (SNV)-genetic algorithm (GA) -the partial least square regression (PLSR) model (master model A) - to predict LC in batch A of snow pear samples based on 80 selected effective wavelengths, with a higher correlation coefficient of prediction set (Rp) of 0.854 and a lower root mean square error of prediction set (RMSEP) of 0.624, which we used as the prediction model to detect LC in three other batches of snow pear samples. The performance of detecting the LC of batch B, C, and D samples by the master model A directly was poor, with lower Rp and higher RMSEP. The independent semi-supervision free parameter model enhancement (SS-FPME) method and the sequential SS-FPME method were used and compared to update master model A to predict the LC of snow pears. For the batch B samples, the predictive ability of the updated model (Ind-model AB) was improved, with an Rp of 0.837 and an RMSEP of 0.614. For the batch C samples, the performance of the Seq-model ABC was improved greatly, with an Rp of 0.952 and an RMSEP of 0.383. For the batch D samples, the performance of the Seq-model ABCD was also improved, with an Rp of 0.831 and an RMSEP of 0.309. Therefore, the updated model based on supervision and learning of new batch samples by the sequential SS-FPME method could improve the robustness and migration ability of the model used to detect the LC of snow pears and provide technical support for the development and practical application of portable detection device.

Physicochemical indicators coupled with multivariate analysis for comprehensive evaluation of matcha sensory quality

The sensory quality of matcha is a pivotal factor in determining consumer acceptance. However, the human sensory panel test is difficult to popularize by virtue of professional requirements and inability to evaluate large samples. The analysis showed that physicochemical indicators of matcha were significantly related to sensory quality. Hence, principal component analysis (PCA) based on selected key physicochemical indicators was proposed to evaluate the sensory quality of matcha in this research. The eight key indicators were selected from twenty-four physicochemical indicators based on least absolute shrinkage and selection operator (LASSO) for the establishment of the PCA comprehensive evaluation model. The results demonstrated that the PCA comprehensive evaluation model achieved superior performance, with −0.895 rc (correlation coefficient in calibration set) and −0.883 rp (correlation coefficient in prediction set) for overall sensory quality. This work demonstrated that LASSO-PCA comprehensive evaluation as an objective protocol has great potential in predicting matcha sensory quality.

Research on Enhanced Detection of Benzoic Acid Additives in Liquid Food Based on Terahertz Metamaterial Devices.

It is very important for human health to supervise the use of food additives, because excessive use of food additives will cause harm to the human body, especially lead to organ failures and even cancers. Therefore, it is important to realize high-sensibility detection of benzoic acid, a widely used food additive. Based on the theory of electromagnetism, this research attempts to design a terahertz-enhanced metamaterial resonator, using a metamaterial resonator to achieve enhanced detection of benzoic acid additives by using terahertz technology. The absorption peak of the metamaterial resonator is designed to be 1.95 THz, and the effectiveness of the metamaterial resonator is verified. Firstly, the original THz spectra of benzoic acid aqueous solution samples based on metamaterial are collected. Secondly, smoothing, multivariate scattering correction (MSC), and smoothing combined with first derivative (SG + 1 D) methods are used to preprocess the spectra to study the better spectral pretreatment methods. Then, Uninformative Variable Elimination (UVE) and Competitive Adaptive Reweighted Sampling (CARS) are used to explore the optimal terahertz band selection method. Finally, Partial Least Squares (PLS) and Least square support vector machine (LS-SVM) models are established, respectively, to realize the enhanced detection of benzoic acid additives. The LS-SVM model combined with CARS has the best effect, with the correlation coefficient of prediction set (Rp) is 0.9953, the root mean square error of prediction set (RMSEP) is 7.3 × 10−6, and the limit of detection (LOD) is 2.3610 × 10−5 g/mL. The research results lay a foundation for THz spectral analysis of benzoic acid additives, so that THz technology-based detection of benzoic acid additives in food can reach requirements stipulated in the national standard. This research is of great significance for promoting the detection and analysis of trace additives in food, whose results can also serve as a reference to the detection of antibiotic residues, banned additives, and other trace substances.

Rapid Determination of Peroxide Value of Peanut Oils During Storage Based on Terahertz Spectroscopy

Terahertz (THz) spectroscopy was used to evaluate peroxide value (PV) of peanut oils during storage. Genetic algorithm (GA) and principal component analysis (PCA) were used as the pretreatment methods to reduce THz high-dimensional data. Interval partial least square (iPLS), back propagation neural network (BPNN), support vector regression (SVR), and partial least squares regression (PLSR) were compared to obtain the best quantitative prediction model. The optimal absorbance spectral region related to peanut oils oxidation was found from 1.1749 to 1.5030 THz by iPLS method. In all chemometric methods used, PCA-SVR showed the best quantitative correlation between absorbance spectra and PV, with the correlation coefficient in prediction set (RP) was 0.9289 and root-mean-square error (RMSEP) was 1.6648 meq O2/kg, respectively. It can be concluded that THz spectroscopy with chemometric methods was an effective tool for rapid determination of PV of peanut oils.

Accurate elemental analysis of alloy samples with high repetition rate laser-ablation spark-induced breakdown spectroscopy coupled with particle swarm optimization-extreme learning machine

High repetition rate laser-ablation spark-induced breakdown spectroscopy (HRR LA-SIBS) coupled with particle swarm optimization-extreme learning machine (PSO-ELM) was developed to realize quantitative elemental analysis of aluminum alloy accurately. A compact Nd: YAG laser operated at 1 kHz was used as the ablation laser and a spark discharge was utilized to enhance the plasma emission. The characteristic lines of the elements were selected as input variables and the parameters of ELM were optimized by the PSO algorithm. The PSO-ELM calibration models were established. The correlation coefficient of prediction set reached to 0.9996, 0.9972 and 0.9999, the root mean square error of prediction set decreased to 0.0138%, 0.0045% and 0.0095% for Mg, Cr, and Cu analysis, respectively. Better predictive performance has been demonstrated in comparison with the univariate and SVM methods. HRR LA-SIBS coupled with PSO-ELM can give convenient, rapid and accurate elemental analysis of alloy samples. Moreover, it will extend the potential applications of HRR LA-SIBS in intelligent manufacturing and on-line monitoring in the future.

Noninvasive human red blood cell counting based on dynamic spectrum

Noninvasive and real-time human red blood cell (RBC) counting has high application value in the process of surgery, basic medical research, disease diagnosis and daily management. In this paper, dynamic spectrum (DS) theory was applied to noninvasive human RBC counting under the guidance of “M + N” theory, meanwhile one-by-one elimination method was used to remove the redundant information and increase the signal-to-noise ratio (SNR) of DS for the first time. To illustrate the effectiveness of this method, DS of 317 volunteers were analyzed. According to the PLS modeling effect, the correlation coefficient of prediction set (Rp) is 0.893, the root mean square error ofprediction set (RMSEP) is 0.196 × 1012/L, the correlation coefficient of calibration set (Rc) is 0.926 and the root mean square error ofcalibration set (RMSEC) is 0.180 × 1012/L. This paper suggests that the DS theory has the potential to realize clinical application of noninvasive human RBC counting.

Highly identification of keemun black tea rank based on cognitive spectroscopy: Near infrared spectroscopy combined with feature variable selection.

From the perspective of combating fraud issues and examining keemun black tea properties, there was a contemporary urgent demand for a keemun black tea rankings identification system. Current rapid evaluation systems had been mainly developed for green tea grade evaluation, but there was space for improvement to establish a highly robust model. The present study proposed cognitive spectroscopy that combined near infrared spectroscopy (NIRS) with multivariate calibration and feature variable selection methods. We defined "cognitive spectroscopy" as a protocol that selects characteristic information from complex spectral data and showed optimal results without human intervention. 700 samples representing keemun black tea from seven quality levels were scanned applying an NIR sensor. To differentiate which wavelength variables of the acquired NIRS data carry key and feature information regarding keemun black tea grades, there were four different variables screening approaches, namely genetic algorithm (GA), successive projections algorithm (SPA), competitive adaptive reweighted sampling (CARS), and shuffled frog leaping algorithm (SFLA), were compared in this study. Cognitive models were developed using least squares support vector machine (LSSVM), back propagation neural network (BPNN) and random forest (RF) methods combined with the optimized characteristic variables from the above variables selection algorithms for the identification of keemun black tea rank quality. Experimental results showed that all cognitive models utilizing the SFLA approach achieved steady predictive results based on eight latent variables and selected thirteen characteristic wavelength variables. The CARS-LSSVM model with the best predictive performance was proposed based on selecting ten characteristic latent variables, and the best performance indicators of the model were as follows: the root mean square error of prediction (RMSEP) was 0.0413, the correlation coefficients of prediction set (Rp) was 0.9884, and the correct discriminant rate (CDR) was 99.01% in the validation process. This study demonstrated that cognitive spectroscopy represented a proper strategy for the highly identification of quality rankings of keemun black tea.

Non-destructive determination of soluble solids content using a multi-region combination model in hybrid citrus

Fruit is complicated natural production and the chemical components are heterogeneous at different position, the comprehensive quality evaluation for intact fruit is thefairest way for avoiding the nonuniform distribution. In order to build a higher accuracy model for soluble solids content (SSC) evaluation, the visible/near infrared spectra of citrus were collected from stem, equator and navel positions to analyze the influence of spectrum measurement position on the prediction accuracy of SSC. The SSC value gradually reduced from navel to equator and stem positions by sequence, but the trend of peel thickness of these three regions were opposite of SSC. The coefficientcorrelation of SSC between intact fruit and three local regions reached remarkablelevel, which proved it was feasible to build an accurate model for SSC prediction of intact fruit using the spectral information of local positions. Then separate local models based on specific position spectrum (stem, equator and navel) were built, the result showed the equator position was more suitable to evaluate SSC of intact fruit than navel and stem positions due to the better prediction accuracy, however the unpredictability and variation of the spectral collection position is a challenge to the prediction ability of intact fruit quality. Next multi-region combination models that fusing spectral information of multiple positions were developed, the combination models of ‘Equator + Navel’ and ‘Stem + Equator + Navel’ achieved optimal performance than other combination models and all separate local models, with the correlation coefficient of prediction set (Rpre) and root mean square error of prediction (RMSEP) of 0.8507, 0.8424 and 0.6015°Brix, 0.5901°Brix, respectively. It indicated that the peel thickness interfered the acquisition of spectral information of flesh layer, but the accuracy and robust of the prediction model could be improved through fusing the spectral information of multiple regions. Therefore, the fusion of multi-information sets should be deserved more attention to build a practicable model that is not sensitive to the variation of spectrum measurement position for SSC evaluation of intact citrus fruit.

A novel hyperspectral microscope imaging technology for rapid evaluation of particle size distribution in matcha

Hyperspectral microscope imaging (HMI) technology as a novel approach was proposed to evaluate physical characteristics of matcha. Particle size distribution as one of the significant physical characteristics was investigated. Data fusion which integrated of textural features from images at 524 nm and key spectral features selected by competitive adaptive reweighed sampling (CARS) were as the raw data for modeling. Models were optimized by cross-validation. Results showed that the performance of models was improved with data fusion. The best ANN models with data fusion were achieved with Rp (correlation coefficient in prediction set) of 0.8020 for D10, 0.8414 for D20, 0.8238 for D30, 0.8124 for D40, 0.8058 for D50, 0.8157 for D60, 0.7643 for D70, 0.7360 for D80, and 0.6313 for D90, respectively. This work demonstrated that HMI technology as a rapid, accurate and high effective protocol has great potential in predicting particle size distribution in matcha powder.

An Optimizing Dynamic Spectrum Differential Extraction Method for Noninvasive Blood Component Analysis.

Dynamic spectra (DS) can greatly reduce the influence of individual differences and the measurement environment by extracting the absorbance of pulsating blood at multiple wavelengths, and it is expected to achieve noninvasive detection of blood components. Extracting high-quality DS is the prerequisite for improving detection accuracy. This paper proposed an optimizing differential extraction method in view of the deficiency of existing extraction methods. In the proposed method, the sub-dynamic spectrum (sDS) is composed by sequentially extracting the absolute differences of two sample points corresponding to the height of the half peak on the two sides of the lowest point in each period of the logarithm photoplethysmography signal. The study was based on clinical trial data from 231 volunteers. Single-trial extraction method, original differential extraction method, and optimizing differential extraction method were used to extract DS from the volunteers' experimental data. Partial least squares regression (PLSR) and radial basis function (RBF) neural network were used for modeling. According to the effect of PLSR modeling, by extracting DS using the proposed method, the correlation coefficient of prediction set (Rp) has been improved by 17.33% and the root mean square error of prediction set has been reduced by 7.10% compared with the original differential extraction method. Compared with the single-trial extraction method, the correlation coefficient of calibration set (Rc) has increased from 0.747659 to 0.8244, with an increase of 10.26%, while the correlation coefficient of prediction set (Rp) decreased slightly by 3.22%, much lower than the increase of correction set. The result of the RBF neural network modeling also shows that the accuracy of the optimizing differential method is better than the other two methods both in calibration set and prediction set. In general, the optimizing differential extraction method improves the data utilization and credibility compared with the existing extraction methods, and the modeling effect is better than the other two methods.

Visualizing distribution of moisture content in tea leaves using optimization algorithms and NIR hyperspectral imaging

Effective visualization of moisture content in tea leaves isveryimportant in the tea cultivation industry and in favor of irrigation management in tea garden. In order to obtain the moisture content distribution map of tea leaves, successive projections algorithm (SPA) coupled with stepwise regression (SPA-SR) and competitive adaptive reweighted sampling (CARS) coupled with stepwise regression (CARS-SR) were proposed to select characteristic wavelengths in this study. The whole region of the tea leaves were selected as region of interest (ROI) to extract the NIR hyperspectral reflectance. Moreover, Savitzky-Golay smoothing (SG), orthogonal signal correction (OSC), multiplicative scatter correction (MSC) and detrending were used to handle with raw spectra. In addition, four feature selection algorithms (SPA, CARS, SPA-SR and CARS-SR) were used to extract the most effective wavelengths. Furthermore, multiple linear regression (MLR) was adopted to establish the prediction models based on spectrum after 20 different combination algorithm treatments. The results showed that SPA-SR and CARS-SR can effectively improve the correlation coefficient of prediction set in established MLR models compared with SPA and CARS, respectively. Besides, the combination algorithm for obtaining the best prediction MLR model was SG-MSC coupled with CARS-SR (Rp2 = 0.8631 and RMSEP = 0.0163), and it was applied to retrieve the distribution of moisture content in tea leaves. Visualizing distribution map of tea leaves offered a more intuitive and comprehensive assessment of moisture contents at each pixel, and it provides a novel approach to evaluate plant irrigation.

Dual-Mean Extraction Method of Dynamic Spectrum for Suppressing Random Noise and Coarse Error

Dynamic spectrum (DS) can greatly reduce the effects of individual differences and measurement environment by extracting the equivalent absorbance of pulsating arterial blood at multiple wavelengths. DS is expected to achieve non-invasive detection of human blood components. This paper proposes a dual-mean extraction method to reduce random noise and coarse error caused by the jitter of the subject in the acquisition process. The “superposition averaging” strategy is adopted twice in this method to reduce the interference of random noise, and a more rigorous standard is used to screen and eliminate coarse error. The single-trial extraction method, the optimizing differential extraction method and the dual-mean extraction method were used to extract the DS from the clinical data of 231 volunteers respectively. The Radial basis function (RBF) neural network was used to model the experimental data of the three extraction results. The difference between the correlation coefficients of calibration set and prediction set was used to evaluate the overfitting degree of the model. The results showed that, the correlation coefficients of prediction set of the data extracted by the dual-mean method reached 0.793, which was higher than the 0.714 of the single-trial method and the 0.749 of the optimizing differential method. And the overfitting degree of the model built by the data of the dual-mean method is the lowest in the next 10 sets of repeated modeling experiments. Compared with the single trial and optimizing differential extraction method, the dual-mean extraction method is better in suppressing random noise and coarse error in DS, and can obtain spectral data with higher signal-to-noise ratio.

Non-destructive prediction of soluble solids content of pear based on fruit surface feature classification and multivariate regression analysis

Improving the prediction accuracy of SSC is the unremitting pursuit in the field of nondestructive optics. The Vis/NIR transmission spectra of ‘Korla’ pear were collected with a portable spectrometer instrument developed by ourselves. In order to study the effects of fruit surface feature such as peel color on determination of SSC and classification of surface locations, three types of SSC prediction models (separate location model, global locations model and average spectra model) and four types of surface location classification models (spectra regions: 550–950, 550–780, 780–950 and 550–700 nm) were built based on the full wavelengths (FWs) and effective wavelengths (EWs), respectively. Results showed that the prediction model of EWs-separate location achieved a best result, the correlation coefficient of prediction set (Rpre) were 0.9408 and 0.9463 for sunlit and shaded side samples, respectively, meanwhile, the classification model based on the 68 EWs selected from 550 to 950 nm achieved an optimal result with the correct classification rate of 97.78% and 96.67% in calibration and prediction sets, respectively. Overall mentioned results above illustrated the fruit surface feature was sensitive to the models of SSC prediction and fruit location classification. Therefore, a compensation model of SSC prediction that is robust and accurate, as well as insensitive to fruit surface feature was built by combining the EWs-classification model with EWs-separate location prediction model, the Rpre and root mean square error of prediction were 0.9368 and 0.5256 °Brix, respectively. In addition, the EWs-global locations model exhibited a negligible effect on the surface feature, although its prediction accuracy had a little inferior to the optimal compensation model. Hence, a complex compensation model with higher prediction accuracy of SSC showed a considerable potential for portable spectrometer instrument.

Prediction of Moisture Content for Congou Black Tea Withering Leaves Using Image Features and Nonlinear Method

Withering is the first step in the processing of congou black tea. With respect to the deficiency of traditional water content detection methods, a machine vision based NDT (Non Destructive Testing) method was established to detect the moisture content of withered leaves. First, according to the time sequences using computer visual system collected visible light images of tea leaf surfaces, and color and texture characteristics are extracted through the spatial changes of colors. Then quantitative prediction models for moisture content detection of withered tea leaves was established through linear PLS (Partial Least Squares) and non-linear SVM (Support Vector Machine). The results showed correlation coefficients higher than 0.8 between the water contents and green component mean value (G), lightness component mean value (L*) and uniformity (U), which means that the extracted characteristics have great potential to predict the water contents. The performance parameters as correlation coefficient of prediction set (Rp), root-mean-square error of prediction (RMSEP), and relative standard deviation (RPD) of the SVM prediction model are 0.9314, 0.0411 and 1.8004, respectively. The non-linear modeling method can better describe the quantitative analytical relations between the image and water content. With superior generalization and robustness, the method would provide a new train of thought and theoretical basis for the online water content monitoring technology of automated production of black tea.

A Nondestructive Real-Time Detection Method of Total Viable Count in Pork by Hyperspectral Imaging Technique

A nondestructive method was developed for assessing total viable count (TVC) in pork during refrigerated storage by using hyperspectral imaging technique in this study. The hyperspectral images in the visible/near-infrared (VIS/NIR) region of 400–1100 nm were acquired for fifty pork samples, and their VIS/NIR diffuse reflectance spectra were extracted from the images. The reference values of TVC in pork samples were determined by classical microbiological plating method. Both partial least square regression (PLSR) model and support vector machine regression model (SVR) of TVC were built for comparative analysis to achieve better results. Different transformation methods and filtering methods were applied to improve the models. The results show that both the optimized PLSR model and SVR model can predict the TVC very well, while the SVR model based on second derivation was better, which achieved with RP (correlation coefficient of prediction set) = 0.94 and SEP (standard error of prediction set) = 0.4570 log CFU/g in the prediction set. An image processing algorithm was then developed to transfer the prediction model to every pixel of the image of the entire sample; the visualizing map of TVC would be displayed in real-time during the detection process due to the simplicity of the model. The results demonstrated that hyperspectral imaging is a potential reliable approach for non-destructive and real-time prediction of TVC in pork.

Nondestructive Measurement of Soluble Solids Content of Kiwifruits Using Near-Infrared Hyperspectral Imaging

This study aims to determine the soluble solids content (SSC) of intact kiwifruits of varieties “Xixuan” and “Huayou” by using near-infrared (NIR) hyperspectral imaging and to investigate which model, developed for a single variety or for two varieties together, had better SSC determination performance. The NIR hyperspectral reflectance images of 200 kiwifruits (100 kiwifruits for each variety) were obtained over the wavelength of 865.11–1,711.71 nm. Mean spectra were extracted from the regions of interest in hyperspectral images of each kiwifruit. The samples were divided into calibration set and prediction set based on the joint x–y distance sample set partitioning method. There were 67 samples in calibration set and 33 samples in prediction set when a single variety was used to establish SSC calibration model, and there were 134 samples (67 “Xixuan” and 67 “Huayou”) in calibration set and 66 samples (33 “Xixuan” and 33 “Huayou”) in prediction set when the two varieties were used together. Successive projections algorithm (SPA) was applied to extract the effective wavelengths (EWs) from full spectra (FS). Nine EWs were selected when a single variety (“Xixuan” or “Huayou”) was used in modeling, and 19 EWs were selected when the two varieties were used together. SSC calibration models were developed based on the partial least squares (PLS) regression and least square support vector machine (LSSVM) modeling methods using the full spectra and extracted EWs as inputs, respectively. The results showed that both calibration and prediction performances of LSSVM models were better than those of PLS. The best SSC determination model for “Xixuan,” “Huayou,” and two varieties together were SPA-LSSVM, FS-LSSVM, and FS-LSSVM with the correlation coefficient of prediction set of 0.766, 0.971, and 0.911, and the root-mean-square error of prediction set of 0.968, 0.589, and 1.137, respectively. The study demonstrates the feasibility of using NIR hyperspectral reflectance imaging technique as a noninvasive method for predicting SSC of kiwifruits and indicates that developing a model for a specific variety is helpful to decreasing prediction error and to improving calculation speed.

A comparative study for improving prediction of total viable count in beef based on hyperspectral scattering characteristics

The objectives of this study were to investigate the feasibility of hyperspectral scattering imaging to predict the bacterial contamination in meat nondestructively, and propose an optimal approach for detecting low levels of total viable count (TVC) contamination in beef. Fresh beef samples were obtained from a commercial slaughtering plant, and stored at 4°C for 0–12days. The visible/near-infrared (VIS/NIR) hyperspectral images in the backscattering mode were acquired from 3–5 beef samples on each day of the experiment, in parallel with microbiological analysis to enumerate the TVC population. Lorentzian function was used to resolve the light scattering information within the hyperspectral image and consequently Lorentzian parameters, which represented different hyperspectral scattering characteristics were extracted. In this study, not only the individual Lorentzian parameters but also the parameter combinations were used to establish the multivariate statistical models for predicting beef TVC, based on the modeling methods of principal component regression (PCR), partial least squares regression (PLSR), and back propagation neural network (BPNN), respectively. The models established using individual Lorentzian parameters did not perform well in predicting low levels of TVC contamination in beef, and the best prediction result could only achieved with the correlation coefficient of prediction set (RP) and root mean squared error of prediction set (RMSEP) of 0.81 and 1.27 log CFU/g, respectively. Based on the parameter combinations, the best modeling results were achieved with RP and RMSEP of 0.86 and 0.93 log CFU/g, 0.87 and 0.79 log CFU/g, 0.90 and 0.88 log CFU/g by PCR, PLSR, and BPNN methods, respectively, which confirmed the superiority of the parameter combination method. The results of this study demonstrated for the first time that hyperspectral scattering imaging combined with Lorentzian function and the proposed parameter combination method could be used to detect low levels of bacterial contamination in beef nondestructively.

Quantitative Analysis of Fish Microbiological Quality Using Electronic Tongue Coupled with Nonlinear Pattern Recognition Algorithms

AbstractThe objective of this study was to establish quantitative evaluation models for fish microbiological quality analysis based on electronic tongue technique coupled with nonlinear pattern recognition algorithms. Crucian carp stored at 4C were used. A commercial electronic tongue system was employed. The total viable counts (TVCs) of fish samples were measured by the classical microbiological plating method. Partial least square regression, support vector regression (SVR) and back propagation neural network (BP‐NN) were applied comparatively to predict TVC values. The multivariate regression models were evaluated by the root mean square error of prediction (RMSEP) and the correlation coefficient in prediction set (Rpre). Results revealed that the performance of BP‐NN model was superior to that of PLS model and SVR model. The RMSEP and Rpre of the BP‐NN model for TVC prediction were 0.211 ln colony‐forming unit (cfu)/g and 0.993, respectively. This study showed that electronic tongue together with BP‐NN model could be a reliable technique for the detection of fish microbiological quality.Practical ApplicationsFish is a highly perishable commodity after harvesting and postmortem as a consequence of microbial breakdown mechanisms. Total viable count (TVC) method is the most widely used microbiological indicator for the evaluation of fish microbiological quality. However, the conventional analytical methods for the determination of TVC are cumbersome and time wasting. This work provides a practical and efficient way for rapid, accurate and convenient determination of TVC in fish using electronic tongue combined with regression algorithms to address these limitations.