Exploring Deep Learning to Predict Coconut Milk Adulteration Using FT-NIR and Micro-NIR Spectroscopy.

Accurately identifying adulterants in agriculture and food products is associated with preventing food safety and commercial fraud activities. However, a rapid, accurate, and robust prediction model for adulteration detection is hard to achieve in practice. Therefore, this study aimed to explore deep-learning algorithms as an approach to accurately identify the level of adulterated coconut milk using two types of NIR spectrophotometer, including benchtop FT-NIR and portable Micro-NIR. Coconut milk adulteration samples came from deliberate adulteration with corn flour and tapioca starch in the 1 to 50% range. A total of four types of deep-learning algorithm architecture that were self-modified to a one-dimensional framework were developed and tested to the NIR dataset, including simple CNN, S-AlexNET, ResNET, and GoogleNET. The results confirmed the feasibility of deep-learning algorithms for predicting the degree of coconut milk adulteration by corn flour and tapioca starch using NIR spectra with reliable performance (R2 of 0.886-0.999, RMSE of 0.370-6.108%, and Bias of -0.176-1.481). Furthermore, the ratio of percent deviation (RPD) of all algorithms with all types of NIR spectrophotometers indicates an excellent capability for quantitative predictions for any application (RPD > 8.1) except for case predicting tapioca starch, using FT-NIR by ResNET (RPD < 3.0). This study demonstrated the feasibility of using deep-learning algorithms and NIR spectral data as a rapid, accurate, robust, and non-destructive way to evaluate coconut milk adulterants. Last but not least, Micro-NIR is more promising than FT-NIR in predicting coconut milk adulteration from solid adulterants, and it is portable for in situ measurements in the future.

Classification of the Crosslink Density Level of Para Rubber Thick Film of Medical Glove by Using Near-Infrared Spectral Data.

Classification of the crosslink density level of para rubber medical gloves by using near-infrared spectral data combined with machine learning is the first time reported in this paper. The spectra of medical glove samples with different crosslink densities acquired by an ultra-compact portable MicroNIR spectrometer were correlated with their crosslink density levels, which were referencely evaluated by the toluene swell index (TSI). The machine learning protocols used to classify the 3 groups of TSI were specified as less than 80% TSI, 80-88% TSI, and more than 88% TSI. The 80-88% TSI group was the group in which the compounded latex was suitable for medical glove production, which made the glove specification comply with the requirements of customers as indicated by the tensile test. The results show that when comparing the algorithms used for modeling, the linear discriminant analysis (LDA) developed by 2nd derivative spectra with 15 k-best selected wavelengths fairly accurately predicted the class but was most reliable among other algorithms, i.e., artificial neural networks (ANN), support vector machines (SVM), and k-nearest neighbors (kNN), due to higher prediction accuracy, precision, recall, and F1-score of the same value of 0.76 and no overfitting or underfitting prediction. This developed model can be implemented in the glove factory for screening purposes in the production line. However, deep learning modeling should be explored with a larger sample number required for better model performance.

Nondestructive Detection of Pesticide Residue (Chlorpyrifos) on Bok Choi (Brassica rapa subsp. Chinensis) Using a Portable NIR Spectrometer Coupled with a Machine Learning Approach.

The contamination of agricultural products, such as vegetables, by pesticide residues has received considerable attention worldwide. Pesticide residue on vegetables constitutes a potential risk to human health. In this study, we combined near infrared (NIR) spectroscopy with machine learning algorithms, including partial least-squares discrimination analysis (PLS-DA), support vector machine (SVM), artificial neural network (ANN), and principal component artificial neural network (PC-ANN), to identify pesticide residue (chlorpyrifos) on bok choi. The experimental set comprised 120 bok choi samples obtained from two small greenhouses that were cultivated separately. We performed pesticide and pesticide-free treatments with 60 samples in each group. The vegetables for pesticide treatment were fortified with 2 mL/L of chlorpyrifos 40% EC residue. We connected a commercial portable NIR spectrometer with a wavelength range of 908-1676 nm to a small single-board computer. We analyzed the pesticide residue on bok choi using UV spectrophotometry. The most accurate model correctly classified 100% of the samples used in the calibration set in terms of the content of chlorpyrifos residue on samples using SVM and PC-ANN with raw data spectra. Thus, we tested the model using an unknown dataset of 40 samples to verify the robustness of the model, which produced a satisfactory F1-score (100%). We concluded that the proposed portable NIR spectrometer coupled with machine learning approaches (PLS-DA, SVM, and PC-ANN) is appropriate for the detection of chlorpyrifos residue on bok choi.

Oscillation effect dataset on the measurement accuracy of load-cell sensor applied to the weigh basket.

The dataset in the form of weight, which will be closely related to the moisture content of agricultural products that have been dried in a mechanical dryer, is important to know precisely. Changes in these properties occur very quickly, so that it is important to prepare a system that is integrated with the mechanical dryer, especially the fluidized dryer type. On the one hand, the fluidizing dryer causes a shock to the weigh basket, connected to the weighing system mechanism. Therefore, this article collects a dataset of the weight of agricultural products (maize and soybeans) that have experienced shocks on two weigh baskets that could potentially be used in fluidization-type mechanical dryers. A load-cell sensor connected to a weigh basket is used to measure the weight of the agricultural product. A new generation of IoT techniques will control the sensor. Its microcontroller will send data to the cloud server via an internet network. There were a total of 120 treatments in the raw dataset. For agriculture engineering researchers, this data will provide benefits in measuring the weight of agricultural material in the form of grain dried in a mechanical dryer, especially the type of fluidized dryer, it can be more accurately explained.

Experimental push and pull force data utilizing self-developed automatic liquid dispensers.

Dataset mechanical properties of an automated liquid dispenser are essential to study for proper design. Therefore, this article includes a push and pull force dataset collected via a load cell sensor on an automatic liquid dispenser self-developed. During one test, nineteen push and pull data were acquired. Measured data is transmitted and saved using internet networks on data cloud servers. The dataset is composed of three types of fluid (i.e., water, soap, and hand sanitizer), three levels of fluid volume (i.e., 50, 150, and 250 ml), and six levels of servo motor rotation angle (i.e., 30°, 60°, 90°, 120°, 150°, 180°). The raw dataset consists of 60 treatments from the 1857 test. This data also provides push and pull force testing of an empty automatic liquid dispenser. The raw data files have been provided. For researchers involved in designing automated liquid dispensers, the dataset may be used to be more reliable in its development. It is possible to prevent over and under design in deciding the energy consumption of an automated liquid dispenser by researching this push and pull force data more deeply. The dataset will be shown as Excel files.

Dataset of adulteration with water in coconut milk using FTIR spectroscopy.

This paper presents the spectroscopic dataset, pre-processing, calibration, and predicted model database of Fourier transform infrared (FTIR) spectroscopy used to detect adulterated coconut milk with water. Absorbance spectral data were acquired and recorded in wavelength range from 2500 to 4000 nm for a total of 43 coconut milk samples. Coconut milk ware prepared in three forms of adulteration. Coconut milk comes from traditional markets and instant coconut milk in Indonesia. Spectra data may also be pre-processed to increase prediction accuracy, robustness performance using normalize, multiplicative scatter correction (MSC), standard normal variate (SNV), 1st derivative, 2nd derivative, and combination of 1st derivative and MSC. Calibration models and cross-validation to forecast those adulteration parameters use two regression algorithms, i.e., principal component regression (PCR) and partial least square regression (PLSR). By looking at its statistical metrics, prediction efficiency can be measured and justified (correlation coefficient (r), correlation of determination (R2), and root mean square error (RMSE)). Obtained FTIR datasets and models can be used as a non-invasive method to predict and determine adulteration on coconut milk.