Root Mean Square Error Of Prediction Research Articles

Application of Multivariate Regression and Artificial Neural Network Modelling for Prediction of Physicochemical Properties of Grape-Skin Compost.

The reusability of by-products in the food industry is consistent with sustainable and greener production; therefore, the aim of this paper was to evaluate the applicability of multiple linear regression (MLR), piecewise linear regression (PLR) and artificial neural network models (ANN) to the prediction of grape-skin compost's physicochemical properties (moisture, dry matter, organic matter, ash content, carbon content, nitrogen content, C/N ratio, total colour change of compost samples, pH, conductivity, total dissolved solids and total colour change of compost extract samples) during in-vessel composting based on the initial composting conditions (air-flow rate, moisture content and day of sampling). Based on the coefficient of determination for prediction, the adjusted coefficient of determination for calibration, the root-mean-square error of prediction (RMSEP), the standard error of prediction (SEP), the ratio of prediction to deviation (RPD) and the ratio of the error range (RER), it can be concluded that all developed MLR and PLR models are acceptable for process screening. Furthermore, the ANN model developed for predicting moisture and dry-matter content can be used for quality control (RER >11). The obtained results show the great potential of multivariate modelling for analysis of the physicochemical properties of compost during composting, confirming the high applicability of modelling in greener production processes.

Soybean polypeptide content detection based on fusion of spectral and image information

Soybean polypeptide is a protein hydrolysate. It has multiple benefits, such as anticancer, antioxidant, etc. Addressing the limitations of traditional methods, such as time-consuming procedures and high costs, this study proposed a novel approach for the rapid, nondestructive detection of soybean polypeptide content. This approach utilizes the fusion of spectral and image information. First, seven preprocessing methods were applied to the spectral data. The results indicated that the model accuracy was highest after applying the standard normal variate preprocessing method. Second, a combined method, consisting of synergy interval partial least squares and iteratively retains informative variables, was proposed to extract characteristic variables. A two-layer Stacking framework ensemble learning model was constructed to compare the detection effects between single spectral input and fusion of spectral and image information input. The results showed that the fusion input model achieved the best performance. The coefficient of determination (R 2) of prediction of the fusion input model reached 0.952, an improvement of 0.061 compared to single spectral input. And root mean square error of prediction (RMSEP) of the fusion input model reached 1.803 × 10−4, a reduction of 7.43 × 10−5 compared to single spectral input. These results indicate that the fusion of spectral and image information can effectively improve the model’s accuracy, offering valuable technical support for soybean breeding and quality detection.

Feasibility of excitation-emission fluorescence spectroscopy in tandem with chemometrics for quantitation of trans-resveratrol in vine-shoot ethanolic extracts.

trans-Resveratrol (TR) is a well-known phytochemical compound with important biological properties. It can be recovered from agri-food by-products or wastes, such as vine shoots. Once recovered, its concentration should be measured, possibly in a green, non-destructive, and efficient manner. With these premises, this work aimed to explore the feasibility of excitation-emission fluorescence spectroscopy combined with chemometrics for the analysis of TR in raw extracts obtained from vine shoots. A total of 75 extracts were produced and analyzed by ultra-performance liquid chromatography method with diode array detection (UPLC-DAD) and spectrofluorimetry. Then, the feasibility of two calibration strategies for TR quantitation was assessed - a parallel factor analysis (PARAFAC)-based calibration and the N-way partial least squares (NPLS) regression. The extracts showed variable TR content, the excitation/emission maxima of which were at around 305/390 nm, respectively. The best PARAFAC-based calibration allowed a root mean square error of prediction (RMSEP) of 22.57 mg L-1 , and a relative prediction deviation (RPD) of 2.91 to be obtained but a large number of PARAFAC components should be considered to improve the predictions. The results of the NPLS regression were slightly better, with a RMSEP of 19.47 mg L-1 , and an RPD of 3.33 in the best case. Fluorescence could be an alternative analytical technique to measure TR in complex samples. Chemometric tools allowed the identification of the TR signal in the fluorescence landscapes, which could be further used for its non-destructive quantitation. The need for a more accurate criterion for optimal PARAFAC complexity emerged. © 2024 Society of Chemical Industry.

Improve the accuracy of FT-NIR for determination of zearalenone content in wheat by using the characteristic wavelength optimization algorithm

The research contributes a unique method to achieve high-precision quantification of zearalenone (ZEN) in wheat, significantly improving accuracy in the analysis. Fourier transform near infrared spectroscopy (FT-NIR) was employed to capture the spectral information of wheat with different mildew degrees. Three feature selection models, competitive adaptive reweighted sampling (CARS), support vector machine-recursive feature elimination (SVM-RFE), and multiple feature-spaces ensemble-least absolute shrinkage and selection operator (MFE-LASSO) were introduced to processed pre-processed near-infrared spectral data and established partial least squares (PLS) regression according to the selected features. The outcomes indicated that the optimal generalization performance was achieved by the PLS model optimized through the MFE-LASSO model. The root mean square error of prediction (RMSEP) was 18.6442 μg·kg-1, coefficient of predictive determination (RP2) was 0.9545, and relative percent deviation (RPD) was 4.3198. According to the results, it is feasible to construct a stoichiometric model for the quantitative determination of ZEN in wheat by using FT-NIR combined with feature selection algorithm, and this method can also be extended to the detection of various molds in other cereals in the future.

Non-Destructive Near-Infrared Technology for Efficient Cannabinoid Analysis in Cannabis Inflorescences.

In the evolving field of cannabis research, scholars are exploring innovative methods to quantify cannabinoids rapidly and non-destructively. This study evaluates the effectiveness of a hand-held near-infrared (NIR) device for quantifying total cannabidiol (total CBD), total delta-9-tetrahydrocannabinol (total THC), and total cannabigerol (total CBG) in whole cannabis inflorescences. Employing pre-processing techniques, including standard normal variate (SNV) and Savitzky-Golay (SG) smoothing, we aim to optimize the portable NIR technology for rapid and non-destructive cannabinoid analysis. A partial least-squares regression (PLSR) model was utilized to predict cannabinoid concentration based on NIR spectra. The results indicated that SNV pre-processing exhibited superior performance in predicting total CBD concentration, yielding the lowest root mean square error of prediction (RMSEP) of 2.228 and the highest coefficient of determination for prediction (R2P) of 0.792. The ratio of performance to deviation (RPD) for total CBD was highest (2.195) with SNV. In contrast, raw data exhibited the least accurate predictions for total THC, with an R2P of 0.812, an RPD of 2.306, and an RMSEP of 1.651. Notably, total CBG prediction showed unique characteristics, with raw data yielding the highest R2P of 0.806. SNV pre-processing emerges as a robust method for precise total CBD quantification, offering valuable insights into the optimization of a hand-held NIR device for the rapid and non-destructive analysis of cannabinoid in whole inflorescence samples. These findings contribute to ongoing efforts in developing portable and efficient technologies for cannabinoid analysis, addressing the increasing demand for quick and accurate assessment methods in cannabis cultivation, pharmaceuticals, and regulatory compliance.

Comparison of optimization algorithms for variable selection to enhance the predictive performance of PLS regression model in determining the concentration of heavy metal Cd in peanut oil

Peanut oil is a widely loved edible vegetable oil in the world. The raw material of peanut oil may be contaminated by heavy metals in the process of planting, processing, and transportation. In this study, the heavy metal Cd content in peanut oil was rapidly determined by FT-NIRS combined with a machine learning model. Three variable optimization algorithms were used for FT-NIRS obtained by experiments, namely, the variable combination population analysis (VCPA), the iteratively variable subset optimization (IVSO), and the multi-feature spaces ensemble strategy by least absolute shrinkage and selection operator (MFE-LASSO). A partial least squares (PLS) regression model was used to determine Cd content in peanut oil. The experimental findings demonstrate that the three variable selection optimization algorithms are effective in identifying key combinations of variables. Among them, the MFE-LASSO-PLS model has strong robustness and generalization performance. Its root mean square error of prediction (RMSEP) was 3.7753 mg·kg-1, prediction coefficient of determination (RP2) was 0.9675, and the relative prediction deviation (RPD) was 5.5485. The research findings demonstrate that the rapid and accurate determination of Cd in peanut oil can be achieved through the integration of FT-NIRS technology with a machine learning model.

Near-Infrared Spectroscopy Modeling of Combustion Characteristics in Chip and Ground Biomass from Fast-Growing Trees and Agricultural Residue

This study focuses on the investigation and comparison of combustion characteristic parameters and combustion performance indices between fast-growing trees and agricultural residues as biomass sources. The investigation is conducted through direct combustion in an air environment using a thermogravimetric analyzer (TGA). Additionally, partial least squares regression (PLSR)-based models were developed to assess combustion performance indices via near-infrared spectroscopy (NIRS), serving as a non-destructive alternative method. The results obtained through the TGA reveal that, specifically, fast-growing trees display higher average ignition temperature (227 °C) and burnout temperature (521 °C) in comparison to agricultural residues, which exhibit the values of 218 °C and 515 °C, respectively. Therefore, fast-growing trees are comparatively difficult to ignite, but sustain combustion over extended periods, yielding higher temperatures. However, despite fast-growing trees having a high ignition index (Di) and burnout index (Df), the comprehensive combustion performance (Si) and flammability index (Ci) of agricultural residue are higher, indicating the latter possess enhanced thermal and combustion reactivity, coupled with improved combustion stability. Five distinct PLSR-based models were developed using 115 biomass samples for both chip and ground forms, spanning the wavenumber range of 3595–12,489 cm−1. The optimal model was selected by evaluating the coefficients of determination in the prediction set (R2P), root mean square error of prediction (RMSEP), and RPD values. The results suggest that the proposed model for Df, obtained through GA-PLSR using the first derivative (D1), and Si, achieved through full-PLSR with MSC, both in ground biomass, is usable for most applications, including research. The model yielded, respectively, an R2P, RMSEP, and RPD, which are 0.8426, 0.4968 wt.% min⁻4, and 2.5; and 0.8808, 0.1566 wt.%2 min⁻2 °C⁻3, and 3.1. The remaining models (Di in chip and ground, Df, and Si in chip, and Ci in chip and ground biomass) are primarily applicable only for rough screening purposes. However, including more representative samples and exploring a more suitable machine learning algorithm are essential for updating the model to achieve a better nondestructive assessment of biomass combustion behavior.

Developing multifruit global near‐infrared model to predict dry matter based on just‐in‐time modeling

AbstractModeling near‐infrared (NIR) spectral data to predict fresh fruit properties is a challenging task. The difficulty lies in creating generalized models that can work on fruits of different cultivars, seasons, and even multiple commodities of fruit. Due to intrinsic differences in spectral properties, NIR models often fail in testing, resulting in high bias and prediction errors. One current solution for achieving generalized models is to use large calibration sets measured over multiple cultivars and harvest seasons. However, current practice primarily focuses on calibration sets for single fruit commodities, disregarding the rich information available from other fruit commodities. This study aims to demonstrate the potential of locally weighted partial least‐squares an example of just‐in‐time (JIT) modeling to develop real‐time models based on calibration sets consisting of multiple fruit commodities. The study also explores JIT modeling for leveraging relevant information from other fruit commodities or adapting the model based on new samples. The application demonstrated here predicts the dry matter in fresh fruit using portable NIR spectroscopy. The results show that JIT modeling is particularly effective for multiple fruit commodities in a single calibration set. The JIT models achieved a root mean squared error of prediction (RMSEP) of 0.69% fresh weight (FW), while the traditional partial least squares (PLS) modeling RMSEP was 0.93% FW. JIT modeling can be particularly beneficial when the user has multiple fruit datasets and wants to combine them into a single dataset to utilize all the relevant information available.

Predicting Silicate Glass Geochemistry Using Raman Spectroscopy and Supervised Machine Learning: Partial Least Square Applications to Amorphous Raman Spectra.

Here, Raman spectroscopy is used to develop a univariate partial least squares (PLS) calibration capable of quantifying geochemistry in synthetic and natural silicate glass samples. The calibration yields eight oxide-specific models that allow predictions of silicon dioxide (SiO2), sodium oxide (Na2O), potassium oxide (K2O), calcium oxide (CaO), titanium dioxide (TiO2), aluminum oxide (Al2O3), ferrous oxide (FeOT), and magnesium oxide (MgO) (wt%) in glasses spanning a wide range of compositions, while also providing correlation-coefficient matrices that highlight the importance of specific Raman channels in the regression of a particular oxide. The PLS suite is trained on 48 of the 69 total glasses, and tested against 21 validation samples (i.e., held out of training). Trends in root mean square error of calibration (RMSEC), root mean square error of cross-validation (RMSECV), and root mean square error of prediction (RMSEP) model accuracy metrics are investigated to uncover the efficacy of utilizing multivariate analysis for such Raman data and are contextualized against recently produced strategies. The technique yields an average root mean of calibration (∼2.4 wt%), cross-validation (∼ 2.9 wt%), prediction (∼ 2.6 wt%), and normalized variance (∼ 28%). Raman band positional shifts are also mapped against underlying chemical variations; with major influences arising primarily as a function of overall oxidation state and silica concentration: via ferric cation (Fe3+)/ferrous cation (Fe2+) ratios and SiO2 (wt%). The algorithm is further validated preliminarily against a separate external set of 11 natural basaltic glasses to unravel the limitations of the synthetic models on natural samples, and to determine the suitability of "universal" Raman-model applications in scenarios where prior chemical contextualization of the target sample is possible. This study represents the first time Raman spectra of amorphous silicates have been paired with PLS, offering a foundation for future improvements utilizing these systems.

In situ rapid evaluation method of quality of peach kernels based on near infrared spectroscopy

This study aimed to perform a rapid in situ assessment of the quality of peach kernels using near infrared (NIR) spectroscopy, which included identifications of authenticity, species, and origins, and amygdalin quantitation. The in situ samples without any pretreatment were scanned by a portable MicroNIR spectrometer, while their powder samples were scanned by a benchtop Fourier transform NIR (FT-NIR) spectrometer. To improve the performance of the in situ determination model of the portable NIR spectrometer, the two spectrometers were first compared in identification and content models of peach kernels for both in situ and powder samples. Then, the in situ sample spectra were transferred by using the improved principal component analysis (IPCA) method to enhance the performance of the in situ model. After model transfer, the prediction performance of the in situ sample model was significantly improved, as shown by the correlation coefficient in the prediction set (Rp), root means square error of prediction (RMSEP), and residual prediction deviation (RPD) of the in situ model reached 0.9533, 0.0911, and 3.23, respectively, and correlation coefficient in the test set (Rt) and root means square error of test (RMSET) reached 0.9701 and 0.1619, respectively, suggesting that model transfer could be a viable solution to improve the model performance of portable spectrometers.

Full Characterisation of Heroin Samples Using Infrared Spectroscopy and Multivariate Calibration.

The analysis of heroin samples, before use in the protected environment of user centra, could be a supplementary service in the context of harm reduction. Infrared spectroscopy hyphenated with multivariate calibration could be a valuable asset in this context, and therefore 125 heroin samples were collected directly from users and analysed with classical chromatographic techniques. Further, Mid-Infrared spectra were collected for all samples, to be used in Partial Least Squares (PLS) modelling, in order to obtain qualitative and quantitative models based on real live samples. The approach showed that it was possible to identify and quantify heroin in the samples based on the collected spectral data and PLS modelling. These models were able to identify heroin correctly for 96% of the samples of the external test set with precision, specificity and sensitivity values of 100.0, 75.0 and 95.5%, respectively. For regression, a root mean squared error of prediction (RMSEP) of 0.04 was obtained, pointing at good predictive properties. Furthermore, during mass spectrometric screening, 10 different adulterants and impurities were encountered. Using the spectral data to model the presence of each of these resulted in performant models for seven of them. All models showed promising correct-classification rates (between 92 and 96%) and good values for sensitivity, specificity and precision. For codeine and morphine, the models were not satisfactory, probably due to the low concentration of these impurities as a consequence of acetylation. For methacetin, the approach failed.

Research on the upgrade and maintenance method of apple soluble solids content models

AbstractIn this paper, 3 years (2020, 2021, and 2022) of Bingtangxin apples were used to study the upgrade and maintenance method of apple soluble solids content (SSC) model using near infrared (NIR) spectroscopy. Modeling by partial least squares (PLS) and one‐dimensional convolutional neural network (1D‐CNN) algorithm. The “base model” is built with apples in 2020 and then upgrade and maintain it. The upgraded model is used to predict the sample in 2021 and 2022. Three methods were used to upgrade and maintenance the model, which are based on updated data, direct correction‐based method, and similar band‐based method. Among the three upgrade methods, the model maintenance method using updated data had the greatest improvement in prediction correlation coefficients (Rp) in both years, and both obtained good predictions. The use of direct correction resulted in the greatest reduction in the root mean square error of prediction (RMSEP) in both years. However, the Rp of the model is not improved much by using the direct correction method. The model maintenance method using similar bands combined with different modeling methods also gives good prediction results. Finally, after comparison, the model maintenance method of selecting similar bands is preferred. In the case of limited improvement in the effect of using the similar band method, the model is maintained using the method of adding updated data. The investigation of model upgrade and maintenance methods can reduce human and material consumption and make apple SSC models more versatile.Practical applicationsNIR technology is the first choice for online fruit quality inspection, but its inspection process requires constant upgrading and maintenance with models. The quality of apples varies from year to year, and the model built with samples from a single year does not work well in predicting samples from other years, resulting in the need to upgrade and maintain the fruit quality detection model. However, sometimes, although a large amount of experimental material is lost, it still results in a generic calibration model that is too cumbersome and leads to poor prediction results. At this time, finding an effective model upgrade and maintenance method can both reduce the sample consumption and make the model prediction accuracy improve, and also reduce the model complexity and make the maintained model more robust.

Monitoring the quality changes and freshness assessment of eggs based on chemometrics method

AbstractDespite their low cost, egg mislabeling is a prevalent problem in the market, resulting in the sale of stale eggs as fresh. In this study, unlike the classical egg freshness parameters such as Haugh Unit and air cell height, UV–Visible (UV‐Vis) spectra, fluorescence spectroscopy, and pH + color properties combined chemometrics methods were used to monitor the eggs' ages. It is known that these methods are able to detect furosine formation and chemical changes during storage. To do this, the qualities of eggs were observed for 15–24 days at three different temperatures: 4°C (refrigeration temperature), 25°C (room temperature), and 35°C (abuse temperature). After conducting a pattern recognition analysis using principal component analysis (PCA), a correlation between storage days and UV–Vis spectra, fluorescence spectra, and pH + color parameters at various temperatures was created using the partial least‐squares (PLS) method. Results showed that PLS models were able to predict eggs' ages successfully. The most reliable age assessment models for eggs were obtained for UV–Vis spectra and fluorescence spectra (ex: 320 nm) of egg samples at 35°C. Real and predicted ages of eggs were presented in validation graphs of UV–Vis spectra and fluorescence spectra (ex: 320 nm) with excellent performance of R2P of 0.96 and 0.97, and root mean square error of prediction (RMSEP) of 1.08 and 0.87, respectively. This method could help food safety organizations use UV–Vis and fluorescence spectrometers at various levels of the egg supply chain as an alternative method.

Determination of zearalenone content in wheat by modified one-dimensional convolutional neural network-based Raman spectra

In this study, a Raman spectral model trained based on Transformer-improved one-dimensional convolutional neural network (1D-CNN) was proposed for the quantitative analysis of zearalenone in wheat. Raman spectra of wheat flour samples with varying levels of mold contamination were acquired using a portable Raman spectrometer. The 1D-CNN was introduced to extract local features of Raman spectra using three convolutional layers, and then the Transformer module was employed to encode and decode these local features for obtaining global information. This approach ensured that the network could better learn complex input spectra data, thereby improving the detection accuracy and precision of the model. The results showed that the improved 1D-CNN model had better generalization performance compared to the 1D-CNN model, with a root mean square error of prediction (RMSEP) of 2.5759 μg/kg, a coefficient of determination (RP2) of 0.9837, and a relative prediction deviation (RPD) of 7.9516. The study demonstrated that the 1D-CNN model improved by Transformer could achieve better feature self-learning and multivariate model correction of Raman spectroscopy data. This research introduces a novel reference method for chemometric analysis of Raman spectra.

Quantitative structure-electrochemistry relationship modeling of a series of anticancer agents using MLR and ANN approaches

Abstract QSPR is a powerful tool for elucidating the correlation between chemical structure and property for both natural and synthesized compounds. In the present work, the half-wave reduction potential for a set of aziridinylquinones (Anticancer Agents [AA]) is modelled using a quantitative structure-electrochemistry relationship (QSER) based on multilinear regression (MLR) and artificial neural network (ANN). Molecular descriptors introduced in this work were computed using the Dragon software (V5). Before the model’s generation, using the Kennard and Stone algorithm, the data set of 84 aziridinylquinones was divided into training and prediction sets consisting of 70 % and 30 % of data points. Quantitative Structure Electrochemistry Relationship (QSER) models were developed using the Genetic Algorithm Multiple Linear Regressions (GA-MLR) and an Artificial Neural Network (ANN). The coefficient of determination (R 2) and Root Mean Squared Error of prediction (RMSE) were mentioned to demonstrate the QSER model’s prediction abilities. Calculated R 2 and RMSEval values for the MLR model were 0.858 and 0.054, respectively. The R 2 and RMSEval values for the ANN training set were calculated to be 0.914 and 0.050, respectively. Findings show that GA is a powerful tool for selecting variables in QSER analysis. Comparing the two employed regression methods showed that ANN is superior to MLR in predictive ability.

Exploration of Global and Specialized Near-Infrared Calibrations for the Quantification of Nutritional Content in Grapevine Organs, Berry Phenological Stages, and Shoot Lignification.

Current infrared spectroscopy applications in the field of viticulture are moving toward direct in-field measuring techniques. However, limited research is available on quantitative applications using direct measurement of fresh tissue. The few studies conducted have combined the spectral data from various cultivars, growing regions, grapevine organs, and phenological stages during model development. The spectral data from these heterogeneous samples are combined into a single data set and analyzed jointly during quantitative analysis. Combining the spectral information of these diverse samples into a global data set could be an unsuitable approach and could yield less accurate prediction results. Spectral differences among samples could be overlooked during model development and quantitative analysis. The development of specialized calibrations should be considered and could lead to more accurate quantitative analyses. This study explored a model optimization strategy attempting global and specialized calibrations. Global calibrations, containing data from multiple organs, berry phenological, and shoot lignification stages, were compared to specialized calibrations per organ or stage. The global calibration for organs contained data from shoots, leaves, and berries and produced moderately accurate prediction results for nitrogen, carbon, and hydrogen. The specialized calibrations per organ yielded more accurate calibrations with a coefficient of determination in validation (R2val) at 90.65% and a root mean square error of prediction (RMSEP) at 0.32% dry matter (DM) for the berries' carbon calibrations. The leaves and shoots carbon calibrations had R2val and RMSEP at 84.99%, 0.34% DM, and 90.06%, 0.37% DM, respectively. The specialized calibrations for nitrogen and hydrogen showed similar improvements in prediction accuracy per organ. Specialized calibrations per phenological and lignification stage were also explored. Not all stages showed improvement, however, most stages had comparable or improved results for the specialized calibrations compared to the global calibrations containing all phenological or lignification stages. The results indicated that both global and specialized calibrations should be considered during model development to optimize prediction accuracy.

Quantitative detection of crude protein in brown rice by near-infrared spectroscopy based on hybrid feature selection

It is important for rice breeding and quality evaluation to predict the protein content of brown rice rapidly and easily. In this study, near-infrared spectroscopy (NIRS) was utilized to establish a model for detecting crude protein content in brown rice based on 349 samples prepared from three kinds of brown rice, and the performance of the model was evaluated. Improved interval partial least squares (iPLS) was used to divide and screen different feature intervals after spectral preprocessing in this research. On this basis, competitive adaptive reweighted sampling (CARS) optimized the selected feature intervals, and finally 14 effective spectral features concentrated in 1160 nm–1338 nm were selected from 1050 features. The above hybrid feature selection has more advantages than the single selection. Support vector regression (SVR) calibration model was established, and partial least squares regression (PLSR) model commonly used in similar studies was selected as a comparison. The optimal spectral preprocessing method was selected according to the model prediction effect. The coefficient of determination (R2), R2 of cross-validation set (Rcv2), root mean square error of prediction (RMSEP), and relative percent difference (RPD) evaluated for the prediction model reached 0.9185, 0.8876, 0.2040% and 3.5194, respectively. The results showed that the designed method can be used for the rapid determination of the crude protein content of brown rice, providing a convenient, efficient and non-destructive method for related detection.

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.

Transfer of near infrared calibration for gasoline octane number based on screening consistent wavelengths combined with direct standardization algorithm

In order to share multivariate calibration models of gasoline research octane number between different near infrared spectrometers, a novel calibration transfer method, namely combination of screening consistent wavelengths and direct standardization (SWCSS-DS) was proposed. Firstly, screening wavelengths with consistent and stable signals (SWCSS) between instruments was used to select the wavelengths with best stability, and then direct standardization (DS) further corrected the systematic errors that still exist after the SWCSS was implemented. The spectra of 120 standard gasoline samples collected on two near infrared spectrometers of the same type were investigated in detail to verify the validity of the new algorithm. Compared results of other transfer methods such as SWCSS, Slope/Bias (S/B), direct standardisation (DS), and piecewise direct standardization (PDS), the root mean squared error for prediction (RMSEP) of SWCSS-DS algorithm for target samples was decreased from 5.75 to 0.295, and the Akaike information criterion value decreased from 1516 to 640, which were better than those of the SWCSS, S/B, DS and PDS algorithms. Therefore, the joint algorithm of SWCSS-DS has not only improved the universality of the master model, but also reduced the dimension of the spectral matrix and calibration equation, that would provide a more efficient model transfer strategy for the practical applications.

Modeling of flaxseed protein, oil content, linoleic acid, and lignan content prediction based on hyperspectral imaging.

Protein, oil content, linoleic acid, and lignan are several key indicators for evaluating the quality of flaxseed. In order to optimize the testing methods for flaxseed's nutritional quality and enhance the efficiency of screening high-quality flax germplasm resources, we selected 30 flaxseed species widely cultivated in Northwest China as the subjects of our study. Firstly, we gathered hyperspectral information regarding the seeds, along with data on protein, oil content, linoleic acid, and lignan, and utilized the SPXY algorithm to classify the sample set. Subsequently, the spectral data underwent seven distinct preprocessing methods, revealing that the PLSR model exhibited superior performance after being processed with the SG smoothing method. Feature wavelength extraction was carried out using the Successive Projections Algorithm (SPA) and the Competitive Adaptive Reweighted Sampling (CARS). Finally, four quantitative analysis models, namely Partial Least Squares Regression (PLSR), Support Vector Regression (SVR), Multiple Linear Regression (MLR), and Principal Component Regression (PCR), were individually established. Experimental results demonstrated that among all the models for predicting protein content, the SG-CARS-MLR model predicted the best, with and of 0.9563 and 0.9336, with the corresponding Root Mean Square Error Correction (RMSEC) and Root Mean Square Error Prediction (RMSEP) of 0.4892 and 0.5616, respectively. In the optimal prediction models for oil content, linoleic acid and lignan, the was 0.8565, 0.8028, 0.9343, and the RMSEP was 0.8682, 0.5404, 0.5384, respectively. The study results show that hyperspectral imaging technology has excellent potential for application in the detection of quality characteristics of flaxseed and provides a new option for the future non-destructive testing of the nutritional quality of flaxseed.