KMITL

Permanent URI for this communityhttps://dspace.kmitl.ac.th/handle/123456789/1

Browse

Search Results

Now showing 1 - 6 of 6

Some of the metrics are blocked by your
consent settings
Quantitative analysis based on image processing combined with machine learning and deep learning to determine the adulteration in nutmeg powder
(2025-10-01)
Sitorus, Agustami
;
Pambudi, Suluh
;
Boodnon, Wutthiphong
;
Lapcharoensuk, Ravipat
The global demand for nutmeg powder has increased, raising the risk of adulteration and necessitating an efficient and cost-effective screening method. The objective of this study is to develop a calibration model to predict the adulteration of nutmeg powder by cinnamon powder using a novel approach by integrating image processing with machine learning (ML) and deep learning (DL). Eight regressors, including four ML regressors (multiple linear ridge regression–MLRR, partial least squares regression–PLSR, multi-layer perceptron–MLP, and adaptive boosting–ABS) and four DL regressors (convolutional neural networks–CNN, AlexNET, residual networks–ResNET, and GoogleNET), were employed to analyze 1800 images of adulteration samples ranging from 0 % to 35 % (w/w). Among ML models, MLP achieved the highest accuracy in prediction (R<inf>p</inf>²=0.922, RMSEP=2.804 %, RPD=3.59), while PLSR (R<inf>p</inf>²=0.876, RMSEP=3.538 %, RPD=2.84), MLRR (R<inf>p</inf>²=0.872, RMSEP=3.596 %, RPD=2.80), and ABS (R<inf>p</inf>²=0.849, RMSEP=3.904 %, RPD=2.58) underperformed. For DL, ResNET (R<inf>p</inf>²=0.882, RMSEP=3.416 %, RPD=2.91) surpassed CNN (R<inf>p</inf>²=0.876, RMSEP=3.505 %, RPD=2.84), AlexNET (R<inf>p</inf>²=0.801, RMSEP=4.429 %, RPD=2.24), and GoogleNET (R<inf>p</inf>²=0.751, RMSEP=4.963 %, RPD=2.00). The MLP's superiority highlights its compatibility with ORB-based feature extraction for nonlinear adulteration patterns, outperforming complex DL architectures. This research highlights the potential of image processing supported by ML and DL as a rapid and low-cost tool for future nutmeg powder adulteration screening.
Some of the metrics are blocked by your
consent settings
Development of automatic tuning for combined preprocessing and hyperparameters of machine learning and its application to NIR spectral data of coconut milk adulteration
(2024-11-01)
Sitorus, Agustami
;
Lapcharoensuk, Ravipat
This study proposed a novel approach to automatically select the preprocessing methods and hyperparameters of machine learning (ML) algorithms based on their best performance in cross-validation for near-infrared (NIR) spectroscopy data. The proposed method simultaneously incorporates single or multiple-preprocessing steps and tunes hyperparameters to determine the best model performance for FT-NIR and Micro-NIR spectral data of coconut milk adulteration with distilled water and mature coconut water in the range of 0%–50%. Computational experiments were conducted using nine single preprocessing types, three types of ML classifier (linear discriminant analysis (LDA), k-nearest neighbour (KNN), multilayer perceptron (MLP)) and three types of ML regressor (partial least squares (PLS), KNN, MLP). The proposed performance strategy effectively addressed and produced satisfactory outcomes for classification and regression challenges in coconut milk adulteration. Finally, the results demonstrated that the proposed approach can more accurately determine the best model, particularly for NIR spectroscopy of coconut milk adulteration.
Some of the metrics are blocked by your
consent settings
Low-Cost Multispectral Sensor for Detecting Adulteration of Onion Powder with Machine Learning
(2024-01-01)
Lapcharoensuk, Ravipat
;
Lunvongsa, Thayanont
;
Suntisakoonwong, Phanchay
;
Sitorus, Agustami
;
Boodnon, Wutthiphong
Onion powder has been frequently adulterated with cheaper materials to boost revenues for scammers. The effective technique is necessary for the application of authentication of onion powder. The aim of this study is application of low-cost multispectral sensor for detecting adulteration of onion powder. Traditional and machine learning algorithms including multiple linear regression (MLR), partial least square regression (PLS-R), nu-support vector regression (nu-SVR), and black propagation neural network (BPNN) were used to train prediction models. Visible and near infrared (Vis-NIR) spectral data was collected using low-cost multispectral sensor at wavelength of 610,680,730, 760,810 and 860 ~nm. Adulterated onion samples were prepared by blending corn flour and onion powder. The coefficient of determination (R_P2) value of all algorithms was between 0.888 and 0.959 while the range of root mean square error of prediction (RMSEP) was 4.214-6.964 %. This finding point indicated that combination of low-cost Vis-NIR multispectral sensor and machine learning could be used for detecting adulteration of onion powder.
Some of the metrics are blocked by your
consent settings
A novel strategy of NIR spectra multivariate calibration in the presence both of small dataset and non-linearity: A comparative study
(2023-12-01)
Devianti
;
Ismy, Adi Saputra
;
Siahaan, Herbert Hasudungan
;
Sitorus, Agustami
The presence of non-linearity and, at the same time, the small number of datasets are often the constraints that appear together in the case of NIR spectra (381–1065nm). This makes some chemometricians think again about presenting a reasonable and robust calibration model using the linear calibration method. On the other hand, even though obtaining a high and robust calibration model, the NIR spectra-based approach is still an alternative method that must still consider low cost and ease of getting it. This study introduces a novel strategy for developing robust calibration models from small and non-linearity NIR spectra datasets. The prediction performance of two groups of chemometric methods, linear (partial least squares regression, PLSR) and non-linear calibration techniques (k-nearest neighbor, k-NN; Ada boosting, AB; Bayesian ridge regression, BRR), were also compared and investigated in depth. A total of forty raw NIR spectral data was used to develop a calibration model to predict the content of B-pinene, D-limonene, and safrole from the nutmeg fruit. The first strategy, non-linearity due to the effect of light scattering on the NIR spectral data, will be handled directly by the non-linear calibration technique algorithm from machine learning to generate the non-linearity model without preprocessing techniques. The second strategy, the robustness of the model, is tested by performing random splitting of data several times without supervision and ending with a rigorous statistical procedure adopted to ensure reliable comparison. The results suggest that the non-linear calibration method is the most promising among the investigated methods. Furthermore, although none of the techniques is always the best to predict on all references, k-NN (for prediction of B-pinene and safrole) and BRR (for prediction of B-pinene and D-limonene), some of them are found to be the most promising in terms of low prediction error (the maximum R<inf>p</inf><sup>2</sup> is 81.6%, and RMSE is less than 1.139%). There are non-linear calibration techniques explored with limited success being achieved.
Some of the metrics are blocked by your
consent settings
An automatic generation of pre-processing strategy combined with machine learning multivariate analysis for NIR spectral data
(2023-09-01)
Arianti, Nunik Destria
;
Saputra, Edo
;
Sitorus, Agustami
Pre-processing near-infrared (NIR) spectral data is indispensable in multivariate analysis, since the measured spectra of complex samples are often subject to overwhelming background, light scattering, varying noises, and other unexpected factors. Various pre-processing methods have been developed to remove or reduce the interference of these effects. Until now, most applications of NIR spectra pre-processing in multivariate calibration have been trial-and-error, with selecting a proper method depending on the nature of the data, expertise, and practitioner experience. Thus, it is usually challenging to determine the best pre-processing method for a given data. In order to tackle these problems, this study proposes a new concept of data pre-processing, namely, automatically generating a pre-processing strategy (AGoES). This concept belongs to the ensemble pre-processing method, where machine learning algorithms (PLSR, SVM, k-NN, DT, AB, and GPR) built on differently preprocessed data are combined by 5-fold cross-validation and grid search optimization. To investigate our concept, a public NIR spectral dataset was used to predict three responses, including dry matter content (DM), organic matter content (OM) and ammonium nitrogen content (AN) from manure organic waste. The results show that SVM is the best algorithm combined with the AGoES pre-processing to predict DM and AN with a ratio of prediction to deviation (RPD) of 3.619 and 2.996, respectively. The AB tandem with AGoES pre-processing is the best strategy for predicting OM with an RPD of 3.185. Therefore, in the framework of the AGoES concept, it is unsupervised pre-processing, more simple, and feasible to apply multivariate analysis using machine learning algorithms.
Some of the metrics are blocked by your
consent settings
Vis-NIR spectra combined with machine learning for predicting soil nutrients in cropland from Aceh Province, Indonesia
(2022-12-01)
Devianti
;
Sufardi
;
Bulan, Ramayanty
;
Sitorus, Agustami
Rapid analytical methods are needed to measure soil nutrient content in cropland, especially in Aceh Province, Indonesia. This is necessary for quick and accurate decision-making on the suitability of the land in terms of soil nutrients and the types of plants to be cultivated on its cropland. Visible near-infrared (Vis-NIR) spectroscopy with suitable chemometric methods through applied machine learning algorithms could be used to predict soil nutrients in the land of agriculture. The current study compared the implementations of machine learning algorithms (support vector machine for regression (SVR), partial least squares artificial neural network (PLS-ANN) and gradient-boosted tree regression (GBRT)) to predict soil nutrients (TN,TP, and TK content) in cropland in Aceh province (Indonesia). The approaches studied used three algorithms of machine learning with four preprocessing employed from spectral data. Samples (n = 102) of soil horizons (0–60 cm) were taken from ten regions in the province of Aceh (Indonesia) and the soil nutrient was measured, including the TN content by the Kjeldahl method and the TP and TK content by the Bray method. Their Vis-NIR spectra (400–2150 nm) were scanned after air drying and ground into powder. 71 examples were used to create the models, while the remaining 31 were used for validation. All of the machine learning algorithms tested as a chemometric approach yielded outstanding models for quantitative estimations of TN, TP, and TK content. Generally, the accuracy of the SVR models of the algorithm utilizing the full spectra was equivalent to that of the PLS-ANN models. Nevertheless, the ANN algorithm using reduced component spectral data (PLS-ANN) served more usefulness than the SVR algorithm depending on the preprocessing method. The most precise models for the content of TN, TP and TK were obtained using the GBRT algorithm (RPD = 2.64, 3.93 and 2.38 for the content of TN, TP and TK, respectively). The results demonstrate that Vis-NIR related to the machine learning algorithm is trustworthy to apply to measure the content of TN, TP, and TK in soil cropland.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results