Pub Date : 2023-07-12DOI: 10.1177/09670335231183086
Zhifeng Chen, Tianhong Pan, Qiong Wu, Xiaofeng Yu
Near infrared (NIR) spectra contain information regarding the analyte as well as uninformative wavelengths. To build high-performance data-driven models, key wavelengths with a strong correlation to the analyte must be selected. This study proposes a feature selection method called stepwise Bayesian linear regression (SBLR) for eliminating unrelated wavelengths, thereby enhancing the robustness of the constructed model. First, a random wavelength is selected from an optimal variable set, and the other wavelengths are placed in a candidate variable set. A Bayesian linear regression (BLR) is implemented by adding a new variable from the candidate set or removing a variable from the optimal set in each step. Furthermore, the BLR model is utilized to perform the F-test. Comparing with the critical value of the F-test with a significance level of α, the test determines whether the variable is retained in the optimal set. Finally, the extracted variables are used to construct a BLR model. The performance and generalization ability of the proposed method were validated. The physical explanation of extracted wavelengths is consistent with the perspective of chemical analysis based on the experiment, which provides a good understanding of the collected NIR spectral data. In addition, compared with traditional algorithms, such as partial least squares regression, least absolute shrinkage and selection operator, and stepwise regression, the proposed method reserves only a few of the effective wavelengths from the full NIR spectra. The proposed method demonstrates potential for key wavelength selection in NIR spectroscopy.
{"title":"Development of feature extraction method for near infrared spectroscopy using stepwise bayesian linear regression","authors":"Zhifeng Chen, Tianhong Pan, Qiong Wu, Xiaofeng Yu","doi":"10.1177/09670335231183086","DOIUrl":"https://doi.org/10.1177/09670335231183086","url":null,"abstract":"Near infrared (NIR) spectra contain information regarding the analyte as well as uninformative wavelengths. To build high-performance data-driven models, key wavelengths with a strong correlation to the analyte must be selected. This study proposes a feature selection method called stepwise Bayesian linear regression (SBLR) for eliminating unrelated wavelengths, thereby enhancing the robustness of the constructed model. First, a random wavelength is selected from an optimal variable set, and the other wavelengths are placed in a candidate variable set. A Bayesian linear regression (BLR) is implemented by adding a new variable from the candidate set or removing a variable from the optimal set in each step. Furthermore, the BLR model is utilized to perform the F-test. Comparing with the critical value of the F-test with a significance level of α, the test determines whether the variable is retained in the optimal set. Finally, the extracted variables are used to construct a BLR model. The performance and generalization ability of the proposed method were validated. The physical explanation of extracted wavelengths is consistent with the perspective of chemical analysis based on the experiment, which provides a good understanding of the collected NIR spectral data. In addition, compared with traditional algorithms, such as partial least squares regression, least absolute shrinkage and selection operator, and stepwise regression, the proposed method reserves only a few of the effective wavelengths from the full NIR spectra. The proposed method demonstrates potential for key wavelength selection in NIR spectroscopy.","PeriodicalId":16551,"journal":{"name":"Journal of Near Infrared Spectroscopy","volume":"31 1","pages":"196 - 203"},"PeriodicalIF":1.8,"publicationDate":"2023-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"48498115","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-07-11DOI: 10.1177/09670335231183104
Li Han, Yan Sun, W. Cai, Xueguang Shao
Near infrared (NIR) spectroscopy has been used to analyze water structures due to the strong absorption of NIR energy by water. The spectral band around 6900 cm−1, corresponding to the first overtone of the OH stretching vibration, is generally studied because the OH in the water molecule with different numbers of hydrogen bonds can be distinguished. In this work, the spectral band around 8600 cm−1, corresponding to the combination of HOH bending and stretching vibration, ν1+ν2+ν3, was studied to extract spectral information about water structures. Continuous wavelet transform was used to enhance the resolution of the spectra. Seven peaks related to the possible molecular structures of water with different numbers of hydrogen bonds were identified based on the spectral changes with temperature. The identification was validated by varying the spectral peaks with molar ratio of H2O–D2O in mixtures and the effect of hydration around the cations on the structure of water. NIR spectroscopy is therefore proven to be a powerful technique for identifying water structures with different hydrogen bonds.
{"title":"Seeking the structure of water from the combination of bending and stretching vibrations in near infrared spectra","authors":"Li Han, Yan Sun, W. Cai, Xueguang Shao","doi":"10.1177/09670335231183104","DOIUrl":"https://doi.org/10.1177/09670335231183104","url":null,"abstract":"Near infrared (NIR) spectroscopy has been used to analyze water structures due to the strong absorption of NIR energy by water. The spectral band around 6900 cm−1, corresponding to the first overtone of the OH stretching vibration, is generally studied because the OH in the water molecule with different numbers of hydrogen bonds can be distinguished. In this work, the spectral band around 8600 cm−1, corresponding to the combination of HOH bending and stretching vibration, ν1+ν2+ν3, was studied to extract spectral information about water structures. Continuous wavelet transform was used to enhance the resolution of the spectra. Seven peaks related to the possible molecular structures of water with different numbers of hydrogen bonds were identified based on the spectral changes with temperature. The identification was validated by varying the spectral peaks with molar ratio of H2O–D2O in mixtures and the effect of hydration around the cations on the structure of water. NIR spectroscopy is therefore proven to be a powerful technique for identifying water structures with different hydrogen bonds.","PeriodicalId":16551,"journal":{"name":"Journal of Near Infrared Spectroscopy","volume":"31 1","pages":"204 - 210"},"PeriodicalIF":1.8,"publicationDate":"2023-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"46007400","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-07-11DOI: 10.1177/09670335231183098
Anja Laubscher, L. Rose, P. Williams
The contamination of maize, a major staple food in South Africa, with fumonisin B1 (FB1), has become a major food safety concern. The regulation of this mycotoxin is extremely important and requires efficient detection methods. Near infrared (NIR) spectroscopy has gained widespread interest as a rapid and non-destructive mycotoxin analysis method. The purpose of this study was, therefore, to determine the NIR absorbance bands of FB1. The spectra of 30 FB1 solutions, constituted in methanol, as well as 30 methanol-only samples were recorded in the spectral range of 1000–2500 nm (10,000 – 4000 cm−1). The data was pre-processed with multiplicative scatter correction (MSC) and a partial least squares discriminant analysis (PLS-DA) model was computed. The variable importance in projection (VIP) scores and selectivity ratio (SR) values were used for wavelength selection. A new PLS-DA model was computed with 454 chosen wavelengths and the regression vector of this model was investigated to further identify and remove irrelevant wavelengths. The final model was computed with 150 wavelengths and nine latent variables (LVs) and obtained a classification accuracy of 100% for both the calibration and external validation sets. By investigating the regression vector of the final PLS-DA model, potential FB1 absorbance bands were identified at 1446 nm, 1453 nm, 1891 nm, 2036 nm, 2046 nm, 2148 nm, 2224 nm, 2262 nm and 2273 nm. This study was therefore able to identify the previously unknown NIR absorbance bands of FB1 at 100 ppm.
{"title":"Determination of potential absorbance bands of fumonisin B1 in methanol with near infrared spectroscopy","authors":"Anja Laubscher, L. Rose, P. Williams","doi":"10.1177/09670335231183098","DOIUrl":"https://doi.org/10.1177/09670335231183098","url":null,"abstract":"The contamination of maize, a major staple food in South Africa, with fumonisin B1 (FB1), has become a major food safety concern. The regulation of this mycotoxin is extremely important and requires efficient detection methods. Near infrared (NIR) spectroscopy has gained widespread interest as a rapid and non-destructive mycotoxin analysis method. The purpose of this study was, therefore, to determine the NIR absorbance bands of FB1. The spectra of 30 FB1 solutions, constituted in methanol, as well as 30 methanol-only samples were recorded in the spectral range of 1000–2500 nm (10,000 – 4000 cm−1). The data was pre-processed with multiplicative scatter correction (MSC) and a partial least squares discriminant analysis (PLS-DA) model was computed. The variable importance in projection (VIP) scores and selectivity ratio (SR) values were used for wavelength selection. A new PLS-DA model was computed with 454 chosen wavelengths and the regression vector of this model was investigated to further identify and remove irrelevant wavelengths. The final model was computed with 150 wavelengths and nine latent variables (LVs) and obtained a classification accuracy of 100% for both the calibration and external validation sets. By investigating the regression vector of the final PLS-DA model, potential FB1 absorbance bands were identified at 1446 nm, 1453 nm, 1891 nm, 2036 nm, 2046 nm, 2148 nm, 2224 nm, 2262 nm and 2273 nm. This study was therefore able to identify the previously unknown NIR absorbance bands of FB1 at 100 ppm.","PeriodicalId":16551,"journal":{"name":"Journal of Near Infrared Spectroscopy","volume":"31 1","pages":"211 - 223"},"PeriodicalIF":1.8,"publicationDate":"2023-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"47312632","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-01DOI: 10.1177/09670335231174328
J. Colling, M. Muller, E. Joubert, F. Marini
Short wave infrared hyperspectral imaging was tested for its ability to distinguish rooibos tea (Aspalathus linearis) based on production area and quality grade, with the aim to replace time-consuming sensory analysis in the industry. The number of latent variables and model parameters of the calibration model were optimised by cross-validation. Classification error rates were used to evaluate the performance of the models in classifying rooibos based on production area and quality grade. The production area of rooibos was distinguished by applying a partial least square-discriminant analysis model with second derivative pre-processing, followed by mean centering and inclusion of nine LVs. The model could successfully distinguish between the two production areas and had a classification accuracy of 100% for the prediction set. To distinguish between different quality grades, a hierarchical model with second derivative pre-processing was developed. Grade A could be distinguished successfully from grades B, C and D (class BCD) with 100% accuracy and grade D could be distinguished from grades B and C (class BC) with 96% accuracy. However, the model was less accurate to distinguish between grade B and C samples, with prediction accuracies of 82 and 66% for B and C, respectively. Application of near infrared hyperspectral imaging therefore offers the potential to replace the use of sensory analysis in the rooibos tea industry to predict production area and quality grade of this herbal tea.
{"title":"Investigating partial least squares discriminant analysis and hierarchical modelling of short wave infrared hyperspectral imaging data to distinguish production area and quality of rooibos (Aspalathus linearis)","authors":"J. Colling, M. Muller, E. Joubert, F. Marini","doi":"10.1177/09670335231174328","DOIUrl":"https://doi.org/10.1177/09670335231174328","url":null,"abstract":"Short wave infrared hyperspectral imaging was tested for its ability to distinguish rooibos tea (Aspalathus linearis) based on production area and quality grade, with the aim to replace time-consuming sensory analysis in the industry. The number of latent variables and model parameters of the calibration model were optimised by cross-validation. Classification error rates were used to evaluate the performance of the models in classifying rooibos based on production area and quality grade. The production area of rooibos was distinguished by applying a partial least square-discriminant analysis model with second derivative pre-processing, followed by mean centering and inclusion of nine LVs. The model could successfully distinguish between the two production areas and had a classification accuracy of 100% for the prediction set. To distinguish between different quality grades, a hierarchical model with second derivative pre-processing was developed. Grade A could be distinguished successfully from grades B, C and D (class BCD) with 100% accuracy and grade D could be distinguished from grades B and C (class BC) with 96% accuracy. However, the model was less accurate to distinguish between grade B and C samples, with prediction accuracies of 82 and 66% for B and C, respectively. Application of near infrared hyperspectral imaging therefore offers the potential to replace the use of sensory analysis in the rooibos tea industry to predict production area and quality grade of this herbal tea.","PeriodicalId":16551,"journal":{"name":"Journal of Near Infrared Spectroscopy","volume":"31 1","pages":"158 - 167"},"PeriodicalIF":1.8,"publicationDate":"2023-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"41365556","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-01DOI: 10.1177/09670335231173136
Xueping Yang, JH Cherney, M. Casler, P. Berzaghi
Portable near infrared (NIR) spectrometers are now readily available on the market and with their smaller size, weight and cost have provided the opportunity to analyze forages both on farms and directly in the field. As new technologies and new portable NIR instruments become available on the market, calibrations for these instruments become a major constraint due to the costs and time necessary to collect reference data. This study evaluated techniques to transfer calibrations for alfalfa and grass forage samples that were developed for a scanning benchtop monochromator (FOSS 6500, 400–2498 nm, LAB) to a diode array instrument (AuroraNir, 950–1650 nm, DA), a digital light processing instrument (NIR-S-G1, 950–1650 nm, DLP) and a short wavelength instrument (SCiO, 740–1070 nm, SCIO). Alfalfa (N = 612) and grass (N = 516) samples from eight agronomic studies were analyzed by wet chemistry for crude protein, neutral detergent fiber (NDF), acid detergent fiber (ADF), in-vitro digestibility (IVTD) and NDF digestibility (NDFD) and divided into calibration, test-set, standardization and inoculation/prediction datasets. Different calibration transfer strategies were evaluated: Spectral Bias Correction (SBC), Shenk and Westerhaus algorithm (SW), Piecewise Direct Standardization (PDS), Dynamic Orthogonal Projection (DOP) or creating a new calibration using LAB predictions of the inoculation/prediction dataset as reference values. All computations for trimming, calibration, validation and standardization were developed using R. SBC with inoculation was an effective method to transfer calibrations for DA. Validation errors for DA transferred calibrations were about 15% lower than LAB for alfalfa data but 6% greater for grass data. For SCIO after DOP spectral adjustment, predicting errors were slightly greater than LAB for both data sets, while prediction errors with DLP were two to three times greater than LAB even after inoculation. PDS created spectral artifacts in the spectra of all three portables, which then resulted in large validation errors. Using LAB predictions as reference values was suitable only for DA, while DLP and DA had large prediction errors. This study showed that calibration sharing between a benchtop and portable instruments is challenging, but possible depending on the portable technologies and the transfer method. Spectral bias correction plus inoculation was the best method to transfer multivariate models for the forage components’ prediction from LAB to handhelds, particularly for DA. Application of DOP was beneficial for SCIO to successfully maintain performance of the original calibration, while for DLP the prediction models were not accurate. Additional studies are necessary to verify these transferring techniques can also be applied to fresh forages, allowing an easier and extended implementation of NIR analysis directly in fields.
{"title":"Forage calibration transfer from laboratory to portable near infrared spectrometers","authors":"Xueping Yang, JH Cherney, M. Casler, P. Berzaghi","doi":"10.1177/09670335231173136","DOIUrl":"https://doi.org/10.1177/09670335231173136","url":null,"abstract":"Portable near infrared (NIR) spectrometers are now readily available on the market and with their smaller size, weight and cost have provided the opportunity to analyze forages both on farms and directly in the field. As new technologies and new portable NIR instruments become available on the market, calibrations for these instruments become a major constraint due to the costs and time necessary to collect reference data. This study evaluated techniques to transfer calibrations for alfalfa and grass forage samples that were developed for a scanning benchtop monochromator (FOSS 6500, 400–2498 nm, LAB) to a diode array instrument (AuroraNir, 950–1650 nm, DA), a digital light processing instrument (NIR-S-G1, 950–1650 nm, DLP) and a short wavelength instrument (SCiO, 740–1070 nm, SCIO). Alfalfa (N = 612) and grass (N = 516) samples from eight agronomic studies were analyzed by wet chemistry for crude protein, neutral detergent fiber (NDF), acid detergent fiber (ADF), in-vitro digestibility (IVTD) and NDF digestibility (NDFD) and divided into calibration, test-set, standardization and inoculation/prediction datasets. Different calibration transfer strategies were evaluated: Spectral Bias Correction (SBC), Shenk and Westerhaus algorithm (SW), Piecewise Direct Standardization (PDS), Dynamic Orthogonal Projection (DOP) or creating a new calibration using LAB predictions of the inoculation/prediction dataset as reference values. All computations for trimming, calibration, validation and standardization were developed using R. SBC with inoculation was an effective method to transfer calibrations for DA. Validation errors for DA transferred calibrations were about 15% lower than LAB for alfalfa data but 6% greater for grass data. For SCIO after DOP spectral adjustment, predicting errors were slightly greater than LAB for both data sets, while prediction errors with DLP were two to three times greater than LAB even after inoculation. PDS created spectral artifacts in the spectra of all three portables, which then resulted in large validation errors. Using LAB predictions as reference values was suitable only for DA, while DLP and DA had large prediction errors. This study showed that calibration sharing between a benchtop and portable instruments is challenging, but possible depending on the portable technologies and the transfer method. Spectral bias correction plus inoculation was the best method to transfer multivariate models for the forage components’ prediction from LAB to handhelds, particularly for DA. Application of DOP was beneficial for SCIO to successfully maintain performance of the original calibration, while for DLP the prediction models were not accurate. Additional studies are necessary to verify these transferring techniques can also be applied to fresh forages, allowing an easier and extended implementation of NIR analysis directly in fields.","PeriodicalId":16551,"journal":{"name":"Journal of Near Infrared Spectroscopy","volume":"31 1","pages":"126 - 140"},"PeriodicalIF":1.8,"publicationDate":"2023-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"45413131","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-01DOI: 10.1177/09670335231173140
Jeremy Walsh, Arjun Neupane, A. Koirala, Michael Li, N. Anderson
The Part 1 prequel to this review evaluated the evolution of modelling techniques used in evaluation of fruit quality over the past three decades and noted a progression towards the use of artificial neural networks (ANNs) and convolutional neural networks (CNNs). In this review, Part 2, the use of CNNs for NIR fruit quality evaluation is explored, given the success of CNNs in various other fields, such as image, video, speech, and audio processing, and the availability of large (open source) datasets of fruit spectra and reference quality attribute, which is required for the training of CNN models. The review provides an overview of deep learning and the CNN architectures and techniques used in NIR spectroscopy for regression modelling, with advantages and disadvantages identified. Studies using CNN for NIR based fruit quality evaluation are then critically examined. Eight publications have presented on models using the same open-source mango dry matter calibration and test set, enabling inter-method comparisons. CNN models have been demonstrated to be accurate, precise and robust. Techniques of transfer learning for CNN models offer an alternative solution to model updating and calibration transfer methods applied in traditional chemometrics. The review has highlighted crucial areas that require resolution and exploration in this application through future research, including, (i) data requirements for training a CNN (ii) optimal spectral pre-processing for CNN (iii) CNN architecture and hyper-parameter selection and tuning for fruit quality evaluation (iv) CNN model interpretability and explainability. Future studies must conduct clearer comparison to partial least squares (PLS) regression and shallow ANNs to better assess the prospective benefit of using CNN, a more complex model. The potential for visualisation of spectra relevance to the CNN model using techniques such as GradCam, currently employed in visualising 2D-CNN models, remains to be explored.
{"title":"Review: The evolution of chemometrics coupled with near infrared spectroscopy for fruit quality evaluation. II. The rise of convolutional neural networks","authors":"Jeremy Walsh, Arjun Neupane, A. Koirala, Michael Li, N. Anderson","doi":"10.1177/09670335231173140","DOIUrl":"https://doi.org/10.1177/09670335231173140","url":null,"abstract":"The Part 1 prequel to this review evaluated the evolution of modelling techniques used in evaluation of fruit quality over the past three decades and noted a progression towards the use of artificial neural networks (ANNs) and convolutional neural networks (CNNs). In this review, Part 2, the use of CNNs for NIR fruit quality evaluation is explored, given the success of CNNs in various other fields, such as image, video, speech, and audio processing, and the availability of large (open source) datasets of fruit spectra and reference quality attribute, which is required for the training of CNN models. The review provides an overview of deep learning and the CNN architectures and techniques used in NIR spectroscopy for regression modelling, with advantages and disadvantages identified. Studies using CNN for NIR based fruit quality evaluation are then critically examined. Eight publications have presented on models using the same open-source mango dry matter calibration and test set, enabling inter-method comparisons. CNN models have been demonstrated to be accurate, precise and robust. Techniques of transfer learning for CNN models offer an alternative solution to model updating and calibration transfer methods applied in traditional chemometrics. The review has highlighted crucial areas that require resolution and exploration in this application through future research, including, (i) data requirements for training a CNN (ii) optimal spectral pre-processing for CNN (iii) CNN architecture and hyper-parameter selection and tuning for fruit quality evaluation (iv) CNN model interpretability and explainability. Future studies must conduct clearer comparison to partial least squares (PLS) regression and shallow ANNs to better assess the prospective benefit of using CNN, a more complex model. The potential for visualisation of spectra relevance to the CNN model using techniques such as GradCam, currently employed in visualising 2D-CNN models, remains to be explored.","PeriodicalId":16551,"journal":{"name":"Journal of Near Infrared Spectroscopy","volume":"31 1","pages":"109 - 125"},"PeriodicalIF":1.8,"publicationDate":"2023-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"47224402","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-26DOI: 10.1177/09670335231173142
Puneet Mishra, Junli Xu
Multimodal measurements are increasingly becoming common in the domain of spectral sensing and imaging for fresh produce. Often multiple sensors are expected to carry complementary information which allows precise estimation of responses. In this study, a novel case of multimodal hyperspectral imaging is described where two different spectral cameras working in the complementary spectral ranges were integrated into a fully standalone system for spectral imaging for fresh produce analysis. Furthermore, a comparative analysis of different multiblock predictive modelling approaches for fusing data from these two complementary spectral cameras is demonstrated. Both multiblock latent space and multiblock variable selection approaches to identify key variables of interest was examined and compared with the analysis carried out on individual data blocks. Prediction of the soluble solids content in grapes was used to demonstrate the application. The presented approach can increase the applications of multimodal hyperspectral imaging for non-destructive analysis.
{"title":"Multimodal close range hyperspectral imaging combined with multiblock sequential predictive modelling for fresh produce analysis","authors":"Puneet Mishra, Junli Xu","doi":"10.1177/09670335231173142","DOIUrl":"https://doi.org/10.1177/09670335231173142","url":null,"abstract":"Multimodal measurements are increasingly becoming common in the domain of spectral sensing and imaging for fresh produce. Often multiple sensors are expected to carry complementary information which allows precise estimation of responses. In this study, a novel case of multimodal hyperspectral imaging is described where two different spectral cameras working in the complementary spectral ranges were integrated into a fully standalone system for spectral imaging for fresh produce analysis. Furthermore, a comparative analysis of different multiblock predictive modelling approaches for fusing data from these two complementary spectral cameras is demonstrated. Both multiblock latent space and multiblock variable selection approaches to identify key variables of interest was examined and compared with the analysis carried out on individual data blocks. Prediction of the soluble solids content in grapes was used to demonstrate the application. The presented approach can increase the applications of multimodal hyperspectral imaging for non-destructive analysis.","PeriodicalId":16551,"journal":{"name":"Journal of Near Infrared Spectroscopy","volume":"31 1","pages":"141 - 149"},"PeriodicalIF":1.8,"publicationDate":"2023-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"42047821","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-22DOI: 10.1177/09670335231173139
P. B. Skou, Margherita Tonolini, C. E. Eskildsen, F. Berg, M. Rasmussen
Partial least squares (PLS) regression is widely used to predict chemical analytes from spectroscopic data, thus reducing the need for expensive and time-consuming wet chemical reference analysis in industrial process monitoring. However, predictions via PLS by definition carry sample-specific errors, and estimation of these errors is essential for correct interpretation of results. To increase trust in PLS regression-based predictions, reliable prediction error estimates must be reported. This can be achieved by determining realistic sample-specific prediction errors using an unbiased mean squared prediction error estimate. This work provides a guide for estimating sample-specific prediction errors, showing the importance of choosing an appropriate error estimator prior to deploying PLS models for industrial applications. We reviewed recent and established methods for estimating the sample-specific prediction error and test them through simulation studies. The methods were subsequently applied for estimating prediction errors in two real-life datasets from the food ingredients industry, where near-infrared spectroscopy was used to quantify i) urea in process water and ii) individual protein concentrations in ultrafiltration retentates from a protein fractionation process. Both the simulations and real data examples showed that the mean squared error of calibration is always a downward biased estimator. Although leave-one-out-cross-validation performed surprisingly well in the data analysed in this work, this paper demonstrated that the appropriate choice of error estimator requires the user to make an informed, data-centered decision.
{"title":"Unbiased prediction errors for partial least squares regression models: Choosing a representative error estimator for process monitoring","authors":"P. B. Skou, Margherita Tonolini, C. E. Eskildsen, F. Berg, M. Rasmussen","doi":"10.1177/09670335231173139","DOIUrl":"https://doi.org/10.1177/09670335231173139","url":null,"abstract":"Partial least squares (PLS) regression is widely used to predict chemical analytes from spectroscopic data, thus reducing the need for expensive and time-consuming wet chemical reference analysis in industrial process monitoring. However, predictions via PLS by definition carry sample-specific errors, and estimation of these errors is essential for correct interpretation of results. To increase trust in PLS regression-based predictions, reliable prediction error estimates must be reported. This can be achieved by determining realistic sample-specific prediction errors using an unbiased mean squared prediction error estimate. This work provides a guide for estimating sample-specific prediction errors, showing the importance of choosing an appropriate error estimator prior to deploying PLS models for industrial applications. We reviewed recent and established methods for estimating the sample-specific prediction error and test them through simulation studies. The methods were subsequently applied for estimating prediction errors in two real-life datasets from the food ingredients industry, where near-infrared spectroscopy was used to quantify i) urea in process water and ii) individual protein concentrations in ultrafiltration retentates from a protein fractionation process. Both the simulations and real data examples showed that the mean squared error of calibration is always a downward biased estimator. Although leave-one-out-cross-validation performed surprisingly well in the data analysed in this work, this paper demonstrated that the appropriate choice of error estimator requires the user to make an informed, data-centered decision.","PeriodicalId":16551,"journal":{"name":"Journal of Near Infrared Spectroscopy","volume":"31 1","pages":"186 - 195"},"PeriodicalIF":1.8,"publicationDate":"2023-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"49568009","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-04-28DOI: 10.1177/09670335231170332
Nuchjira Jindagul, Yuranan Bantadjan, M. Chamchong
The main goal of this study was to predict the age-after-harvest of milled rice and classify it for stale or fresh rice during storage by determining the thiobarbituric acid (TBA) value non-destructively via a hyperspectral imaging (HSI). Thai jasmine rice (KDML 105 variety) was stored at 25°C, 35°C, and 50°C and randomly sampled every month for 12 months for TBA testing (for 4 months at 50°C). During storage, the chemical analysis value of TBA increased over the storage time at all storage temperatures. Hyperspectral imaging in the range 864–1695 nm was used, and partial least squares regression was used to develop multivariate calibration models. The resulting prediction model could approximate quantitative values for TBA with a ratio of performance to the deviation at 2.0 and the root mean square error of prediction of 3.20 μmol MDA/kg. Partial least squares discriminant analysis was conducted for quality analysis based on the TBA value. The age-after-harvest prediction model and the model for classifying stale or fresh rice effectively performed on milled rice, providing a total cross-validation accuracy of 98% and 100%, respectively.
{"title":"Use of hyperspectral chemical imaging to determine the age of milled rice post harvest","authors":"Nuchjira Jindagul, Yuranan Bantadjan, M. Chamchong","doi":"10.1177/09670335231170332","DOIUrl":"https://doi.org/10.1177/09670335231170332","url":null,"abstract":"The main goal of this study was to predict the age-after-harvest of milled rice and classify it for stale or fresh rice during storage by determining the thiobarbituric acid (TBA) value non-destructively via a hyperspectral imaging (HSI). Thai jasmine rice (KDML 105 variety) was stored at 25°C, 35°C, and 50°C and randomly sampled every month for 12 months for TBA testing (for 4 months at 50°C). During storage, the chemical analysis value of TBA increased over the storage time at all storage temperatures. Hyperspectral imaging in the range 864–1695 nm was used, and partial least squares regression was used to develop multivariate calibration models. The resulting prediction model could approximate quantitative values for TBA with a ratio of performance to the deviation at 2.0 and the root mean square error of prediction of 3.20 μmol MDA/kg. Partial least squares discriminant analysis was conducted for quality analysis based on the TBA value. The age-after-harvest prediction model and the model for classifying stale or fresh rice effectively performed on milled rice, providing a total cross-validation accuracy of 98% and 100%, respectively.","PeriodicalId":16551,"journal":{"name":"Journal of Near Infrared Spectroscopy","volume":"31 1","pages":"150 - 157"},"PeriodicalIF":1.8,"publicationDate":"2023-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"44363576","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-02-20DOI: 10.1177/09670335231156471
Yongli Bai, Xinguo Huang, Nan Peng, S. Zhang, Yunfei Zhong
Water-based inks are widely used in green packaging and printing. The printability parameters of water-based inks, such as viscosity (alcohol concentration (AC)) and color (toning additive concentration (toning yellow concentration/toning red concentration, TYC/TRC)), can only be controlled manually in many printing companies. The printability parameters of water-based inks with different additives were analyzed using spectral preprocessing, variable selection, and model-building methods with visible and near infrared (vis-NIR) spectral data (380∼980 nm). Model performance was compared using the root mean square error of cross-validation (RMSEC) and the coefficient of determination (R2). The results of the experiment indicate that the viscosity of the water-based inks can be quantitatively predicted using the principal component analysis and back propagation neural network model (PCA-BPNN) combined with Savitzky-Golay (SG) smoothing in the spectral subrange, which is superior to the PLS regression model. The R2c and r2p of the PCA-BPNN model were up to 0.998 and 0.993, and the RMSEC and RMSEP values obtained were 0.21 and 0.34. Similarly, the concentration of toning yellow and toning red can be quantitively predicted using the PCA-BPNN model combined with SG smoothing in the 617∼726 nm spectral range, which is better than iPLS regression model. These results indicate that the use of vis-NIR spectroscopy and chemometrics is a promising strategy, reliable for predicting the printability parameters of water-based inks, and provides the technical basis for subsequent implementation of online inspection.
{"title":"Development of a quantitative method to evaluate the printability parameters of water-based ink using visible and near infrared spectroscopy","authors":"Yongli Bai, Xinguo Huang, Nan Peng, S. Zhang, Yunfei Zhong","doi":"10.1177/09670335231156471","DOIUrl":"https://doi.org/10.1177/09670335231156471","url":null,"abstract":"Water-based inks are widely used in green packaging and printing. The printability parameters of water-based inks, such as viscosity (alcohol concentration (AC)) and color (toning additive concentration (toning yellow concentration/toning red concentration, TYC/TRC)), can only be controlled manually in many printing companies. The printability parameters of water-based inks with different additives were analyzed using spectral preprocessing, variable selection, and model-building methods with visible and near infrared (vis-NIR) spectral data (380∼980 nm). Model performance was compared using the root mean square error of cross-validation (RMSEC) and the coefficient of determination (R2). The results of the experiment indicate that the viscosity of the water-based inks can be quantitatively predicted using the principal component analysis and back propagation neural network model (PCA-BPNN) combined with Savitzky-Golay (SG) smoothing in the spectral subrange, which is superior to the PLS regression model. The R2c and r2p of the PCA-BPNN model were up to 0.998 and 0.993, and the RMSEC and RMSEP values obtained were 0.21 and 0.34. Similarly, the concentration of toning yellow and toning red can be quantitively predicted using the PCA-BPNN model combined with SG smoothing in the 617∼726 nm spectral range, which is better than iPLS regression model. These results indicate that the use of vis-NIR spectroscopy and chemometrics is a promising strategy, reliable for predicting the printability parameters of water-based inks, and provides the technical basis for subsequent implementation of online inspection.","PeriodicalId":16551,"journal":{"name":"Journal of Near Infrared Spectroscopy","volume":"31 1","pages":"100 - 106"},"PeriodicalIF":1.8,"publicationDate":"2023-02-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"42493831","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：