Hyperspectral Imaging with Machine Learning Approaches for Assessing Soluble Solids Content of Tribute Citru

Li, Cheng; He, Mengyu; Cai, Zeyi; Qi, Hengnian; Zhang, Jianhong; Zhang, Chu

doi:10.3390/foods12020247

Cited by 14 publications

(5 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…By training the models with a broader range of spectral information, the variations attributed to sample preparation and Vis-NIRs analysis would be mitigated. It should be noted that there are studies based on spectroscopy data in which better results were obtained with PLS than with SVR [ 53 ], while in other cases, the opposite is reported [ 54 ] or even similar performances [ 55 ].…”

Section: Resultsmentioning

confidence: 99%

Rapid and Automated Method for Detecting and Quantifying Adulterations in High-Quality Honey Using Vis-NIRs in Combination with Machine Learning

Calle,

Punta-Sánchez,

González-de-Peredo

et al. 2023

Foods

View full text Add to dashboard Cite

Honey is one of the most adulterated foods, usually through the addition of sweeteners or low-cost honeys. This study presents a method based on visible near infrared spectroscopy (Vis-NIRs), in combination with machine learning (ML) algorithms, for the correct identification and quantification of adulterants in honey. Honey samples from two botanical origins (orange blossom and sunflower) were evaluated and adulterated with low-cost honey in different percentages (5%, 10%, 15%, 20%, 25%, 30%, 35%, 40%, 45%, and 50%). The results of the exploratory analysis showed a tendency to group the samples according to botanical origin, as well as the presence of adulteration. A supervised analysis was performed to detect the presence of adulterations. The best performance with 100% accuracy was achieved by support vector machines (SVM) and random forests (RF). A regression study was also carried out to quantify the percentage of adulteration. The best result was obtained by support vector regression (SVR) with a coefficient of determination (R2) of 0.991 and a root mean squared error (RMSE) of 1.894. These results demonstrate the potential of combining ML with spectroscopic data as a method for the automated quality control of honey.

show abstract

Section: Resultsmentioning

confidence: 99%

Rapid and Automated Method for Detecting and Quantifying Adulterations in High-Quality Honey Using Vis-NIRs in Combination with Machine Learning

Calle,

Punta-Sánchez,

González-de-Peredo

et al. 2023

Foods

View full text Add to dashboard Cite

show abstract

“…In each iteration, the wavelength contributing the most to the model, based on correlation, is selected and added to the subset. This process effectively reduces dimensionality by eliminating multicollinear and redundant variables using SPA [30][31][32].…”

Section: Spectral Preprocessing and Model Developmentmentioning

confidence: 99%

Effect of Combined Non-Wood and Wood Spectra of Biomass Chips on Rapid Prediction of Ultimate Analysis Parameters Using near Infrared Spectroscopy

Shrestha,

Posom,

Sirisomboon

et al. 2024

Energies

View full text Add to dashboard Cite

The ultimate analysis parameters, including carbon (C), hydrogen (H), nitrogen (N), and oxygen (O) content in biomass, were rarely found to be predicted by non-destructive tests to date. In this research, we developed partial least squares regression (PLSR) models to predict the ultimate analysis parameters of chip biomass using near-infrared (NIR) raw spectra of non-wood and wood samples from fast-growing tree and agricultural residue and nine different traditional spectral preprocessing techniques. These techniques include first derivative (sd1), second derivative (sd2), constant offset, standard normal variate (SNV), multiplicative scatter correction (MSC), vector normalization, min-max normalization, mean centering, sd1 + vector normalization, and sd1 + MSC. Additionally, we employed a genetic algorithm (GA), successive projection algorithm (SPA), multi-preprocessing (MP) 5-range, and MP 3-range to develop a PLSR model for rapid prediction. A dataset consisting of 120 chip biomass samples was utilized for model development in which the samples were non-wood samples of 65–67% and wood samples of 33–35%, and the model performance was evaluated and compared. The selection of the optimum performing model was mainly based on criteria such as the coefficient of determination in the prediction set (R2P), root mean square error of the prediction set (RMSEP), and the ratio of prediction to deviation (RPD). The optimal model for weight percentage (wt.%) of C was obtained using GA–PLSR, yielding R2P, RMSEP, and RPD values of 0.6954, 1.1252 wt.%, and 1.8, respectively. Similarly, for wt.% of O, the most effective model was obtained using the multi-preprocessing PLSR–5 range method with R2P of 0.7150, RMSEP of 1.3088 wt.%, and RPD of 1.9. For wt.% of N, the optimal model was obtained using the MP PLSR-3 range method, resulting in R2P, RMSEP, and RPD values of 0.6073, 0.1008 wt.%, and 1.6, respectively. However, wt.% of the H model provided R2P, RMSEP, and RPD values of 0.5162, 0.2322 wt.%, and 1.5, respectively. Notably, the limit of quantification (LOQ) values for C, H, and O were lower than the minimum reference values used during model development, indicating a high level of sensitivity. However, the LOQ for N exceeded the minimum reference value, implying the samples to be predicted by the model must be in the range of reference range in the calibration set. By scatter plot analysis, the effect of combined non-wood and wood spectra of biomass chips on rapid prediction of ultimate analysis parameters using NIR spectroscopy was investigated. To include different species in a model, the species have to be not only in the different values of the constituents to make a wider range for a robust model, but also must provide their trend line characteristics in the scatter plot, i.e., correlation coefficient (R), slope, and intercept (same slope and slope approached to 1, and intercept is same (no gap) and approached zero, high R approached to 1). The effect of the R, slope, and intercept to obtain the better-optimized model was studied. The results show that the different species affected the model performance of each parameter prediction in a different manner, and by scatter plot analysis, which of these species were affecting the model negatively and how the model could be improved was indicated. This is the first time the effect has been studied by the principle of a scatter plot.

show abstract

“…The subsequent step assesses these candidate subsets based on the RMSE values obtained from the validation set of the PLSR calibration. The final step involves removing uninformative variables through a variable elimination procedure that does not significantly compromise predictive ability [42][43][44].…”

Section: Partial Least Squares Regression Model Developmentmentioning

confidence: 99%

NIR Spectroscopy as an Alternative to Thermogravimetric Analyzer for Biomass Proximate Analysis: Comparison of Chip and Ground Biomass Models

Shrestha,

Posom,

Sirisomboon

et al. 2024

Energies

View full text Add to dashboard Cite

This study investigates the non-destructive analysis of proximate parameters (moisture content, MC; volatile matter, VM; fixed carbon, FC; ash content) in various chipped and ground biomass using a combination of destructive thermogravimetric analysis (TGA) and non-destructive near-infrared spectroscopy (NIRS) with partial least squares regression (PLSR). The thermogravimetric method determines proximate analysis data through TG and DTG curves, tracking biomass mass loss over time or temperature. NIRS scans chipped biomass in diffuse reflectance, and ground biomass in transflectance mode, covering the wavenumber range from 3595 to 12,489 cm−1. PLSR-based models (Full-PLSR, GA-PLSR, SPA-PLSR, MP PLSR 5-range method, and MP PLSR 3-range method) are developed and evaluated based on R2P, RMSEP, and RPD. MC and FC models for chip biomass exhibit satisfactory performance, making them cautiously applicable in various applications, including research. Optimal models for MC and FC in chip biomass, constructed using GA-PLSR with the second derivative and Full-PLSR with a constant offset, yield high R2P values (0.8654 and 0.8773), low RMSEP values (0.85% and 2.12%), and high RPD values (2.9 and 3.0), indicating applicative capabilities. Other parameters such as MC and FC in ground biomass, as well as VM and ash content in both chip and ground biomass, are found suitable for rough screening. Model sensitivity, assessed by calculating LOQ, indicates high sensitivity for VM in both chip and ground biomass and FC in chip biomass, as the calculated LOQ value is lower than the minimum reference values used during model development. However, for the remaining parameters, LOQ values surpass the established minimum reference value, suggesting limitations in predicting samples below the calibration range. Continuous model enhancement incorporating an ample number of representative biomass samples and consistent validation with unknown samples are imperative for ensuring accurate predictions.

show abstract

Hyperspectral Imaging with Machine Learning Approaches for Assessing Soluble Solids Content of Tribute Citru

Cited by 14 publications

References 32 publications

Rapid and Automated Method for Detecting and Quantifying Adulterations in High-Quality Honey Using Vis-NIRs in Combination with Machine Learning

Rapid and Automated Method for Detecting and Quantifying Adulterations in High-Quality Honey Using Vis-NIRs in Combination with Machine Learning

Effect of Combined Non-Wood and Wood Spectra of Biomass Chips on Rapid Prediction of Ultimate Analysis Parameters Using near Infrared Spectroscopy

NIR Spectroscopy as an Alternative to Thermogravimetric Analyzer for Biomass Proximate Analysis: Comparison of Chip and Ground Biomass Models

Contact Info

Product

Resources

About