Skip to main content
NCBI home page
Food Chemistry: X logoLink to Food Chemistry: X
. 2026 Jul 8;38:104193. doi: 10.1016/j.fochx.2026.104193

Rapid and non-destructive estimation of chemical properties of date syrup using visible-near infrared hyperspectral imaging combined with machine learning predictive models

Mohammad Hossein Nargesi 1,⁎, Kamran Kheiralipour 1,⁎
PMCID: PMC13401031  PMID: 42502869

Abstract

In this study, an innovative hybrid approach combining hyperspectral imaging and artificial intelligence was developed to replace costly and time-consuming laboratory methods of determining chemical properties of date syrup. Hyperspectral data related to acidity, Brix, fructose, HMF, pH, and viscosity were analyzed using three machine learning models: Artificial neural network, partial least squares regression, and support vector regression. These models were evaluated under two strategies, with and without feature selection step. Among the models, the Artificial Neural Network demonstrated superior performance compared to the other methods. The results revealed that the optimal performance was achieved with four neurons in the hidden layer and the use of the Levenberg-Marquardt activation function. The prediction accuracies for the mentioned properties were 100%, 99.99%, 100%, 99.99%, 100%, and 99.99%, respectively. In the strategy without feature selection, the prediction accuracy for all chemical parameters reached 100%. These findings highlight the capability of hyperspectral imaging and artificial neural networks in estimating chemical properties of date syrup.

Keywords: Hyperspectral imaging, Image analysis, Machine learning, Quality, Date syrup

Highlights

  • •

    Introduced a non-destructive method for predicting chemical parameters in date sap.

  • •

    Presented spectral data from an HSI device.

  • •

    The efficient features were extracted from effective channels and then classified.

  • •

    An artificial neural network approach was used to predict the chemical parameters of date sap.

  • •

    The prediction accuracy for different levels of date sap chemical parameters was reported 100%.

1. Introduction

The date palm (Phoenix dactylifera L.) is considered one of the most important agricultural crops in many Middle Eastern countries, with its fruit-dates-being a staple in the diet of millions of people (Alkhoori et al., 2022). Dates are an excellent source of energy, containing approximately 78% carbohydrates, 2–3% protein, and less than 1% fat. In addition to being a natural energy source, they play a key nutritional role due to their significant content of vitamins, minerals, and dietary fiber (El-Nagga & Abd El–Tawab, 2012). Dates are considered a highly nutritious fruit due to their rich content of carbohydrates, proteins, and essential minerals. They are an excellent source of thermal energy for the human body, with sugars making up approximately 80% of their total weight (Shehzad et al., 2021; Siddeeg et al., 2019; Siddiq et al., 2013). The high potassium content and extremely low sodium levels make dates an ideal dietary choice for individuals with high blood pressure (Keshtkaran & Mohammadifar, 2013). Epidemiological studies have consistently demonstrated that regular and high consumption of dates is associated with a reduced risk of chronic conditions such as cardiovascular diseases, various types of cancer, aging, atherosclerosis, neurodegenerative disorders, tumors, and mutagenic agents (Al-Farsi et al., 2007; El-Beltagy et al., 2009; Gad et al., 2010). Furthermore, research indicates that consuming date extracts during the postpartum period can enhance the immune system in women (Gad et al., 2010). In addition, a polysaccharide isolated from dates has shown notable antitumor properties (Ishurd & Kennedy, 2005). Furthermore, dates are rich in a wide range of bioactive compounds that exhibit antioxidant, antimicrobial, anticancer, anti-inflammatory, and even anti-infertility properties (Alhamdan et al., 2021). Industrially, dates are used to produce a variety of products such as syrup, juice, jam, jelly, candy, processed dates, pickles, and vinegar. Among these, date syrup is one of the most widely consumed derivatives. It is highly valued for its rich content of glucose, fructose, minerals, and organic acids. Due to the presence of polyphenols and similar compounds, date syrup also exhibits notable antioxidant activity (Alhamdan et al., 2021; El-Nagga & Abd El–Tawab, 2012). It is widely used in the preparation of both traditional and industrial foods, such as beverages, confectionery, ice cream, and dairy-based desserts like yogurt (Gholamhosseinpour & Hashemi, 2022; Keshtkaran & Mohammadifar, 2013).

Due to their nutritional composition, low-quality dates have also been used to produce value-added products. Date syrup, known locally in some countries as “Rub Al Tamr,” is prepared from specific local varieties of dates in nations such as Libya, Iraq, and Tunisia (Barreveld, 1993). This syrup is either consumed directly or used as an ingredient in the formulation of food products such as ice cream, beverages, sweets, bakery items, sesame paste/date syrup blends, jams, and date butter (Barreveld, 1993; Razavi et al., 2007; Roukas & Kotzekidou, 1997). Date syrup is a processed product with high nutritional value, obtained as a thick liquid from dates through thermal extraction. This syrup is free from fibers, sediments, impurities, and foreign matter, making it suitable for direct consumption or use in the production of confectionery and bakery products (Yahmed et al., 2021). It consists of 86.6% total sugars, 81.7% reducing sugars, and 4.9% sucrose. Additionally, it contains 24.8% moisture, 0.02% acidity, 2.1% protein, and 6.6% ash on a dry weight basis. The syrup is also a good source of vitamins A and B (Ashraf & Hamidi-Esfahani, 2011) and is rich in minerals (Al Hilfi et al., 2019). Date syrup is a valuable food source, rich in sugars, minerals, vitamins, and antioxidants (Al-Farsi et al., 2018). Research has shown that dates have a low glycemic index, making them a beneficial option whether consumed alone or in combination with plain yogurt (Yousif et al., 1996). Date fruit exhibits tumor-fighting properties (Ishurd & Kennedy, 2005), as well as antioxidant, anticancer, and antimutagenic effects (Maqsood et al., 2020). However, date syrup extracted through traditional methods often appears darker in color due to the prolonged application of high heat, which leads to the formation of colored compounds such as caramel, resulting from reactions between sugars and amino acids. Another common issue is crystallization, caused by the separation and accumulation of sugars, which can be managed by adding citric acid. Reduced fluidity is also a concern, as the presence of pectic substances gives the syrup a jelly-like consistency. Furthermore, if the concentration of the syrup drops below 75%, there is an increased risk of fermentation and microbial contamination due to insufficient hygiene (Ashraf & Hamidi-Esfahani, 2011). Quality assessment of materials is a fundamental component in achieving sustainable production (Kheiralipour, 2022). The evaluation of physicochemical properties of date syrup such as hydroxymethylfurfural (HMF) content, Brix value, fructose concentration, moisture content, pH, acidity, and viscosity are typically conducted using advanced laboratory instruments under the supervision of skilled operators (Caliskan & Polat, 2011; Cepeda & Villaran, 1999; Farahnaky et al., 2016). However, these procedures are time-consuming, costly, and require access to specialized laboratory facilities. Therefore, the development and application of rapid, non-destructive, and accurate methods and also combined with machine learning algorithms can play a crucial role in quality monitoring of food and optimization of production processes (Nargesi et al., 2026; Deng et al., 2026). Different imaging techniques are avilable to record area data of products with high speed and accuracy and low cost (Kheiralipour, 2012). Hyperspectral imaging (HSI) is an advanced technique that integrates spectroscopy with imaging sensors to simultaneously capture high-resolution spatial and spectral information (Kheiralipour, 2024). This technique serves as a powerful tool for identifying and analyzing the chemical and physical properties of food products and related materials (He & Sun, 2015; Wang et al., 2021; Kheiralipour, Singh, & Jayas, 2023; Sun et al., 2024; Ahmed et al., 2024). This technique generates images containing comprehensive spectral data for each pixel, allowing detailed spectral characterization of samples (Kheiralipour, Chelladurai, & Jayas, 2023). Unlike conventional imaging systems that rely solely on three broad color bands (RGB), HSI acquires data across hundreds of narrows, contiguous spectral bands over a wide wavelength range (Kamruzzaman, 2023; Kheiralipour & Jayas, 2024). This broad spectral coverage facilitates the extraction of specific information regarding chemical composition, structural features, and various physical attributes of food samples (Elmasry et al., 2012; Jia et al., 2024; Kheiralipour, Sajadipour, & Nadimi, 2025; Nargesi, Kheiralipour, & Khodaei, 2025). The quality and utility of this hyperspectral data are inherently influenced by sensor characteristics such as signal-to-noise ratio, spectral resolution, and spatial resolution; recent research has explored how evolving artificial intelligence techniques can adapt to and leverage these variable data qualities for improved analysis (Li, Liu, Cheng, et al., 2026; Li, Liu, Huang, et al., 2026). One of the key advantages of HSI is its non-destructive and non-invasive nature, which allows for in-depth analysis without altering or damaging the sample (Kheiralipour, Sajadipour, & Nargesi, 2025). This feature is particularly important for food assessment, as it preserves product integrity, avoids sample preparation steps that may affect its physicochemical properties, and enables real-time monitoring during industrial processing and quality control operations (Kamruzzaman et al., 2015). In addition to its applications in the food industry, HSI is widely employed in environmental monitoring tasks such as detecting vegetation changes, assessing forest health, and tracking land-use transformations (Lee et al., 2024). Furthermore, recent advancements, such as the multi-scale autoencoder suppression strategy for hyperspectral image anomaly detection, have significantly enhanced the capability to identify spectral irregularities (Tu et al., 2025). Within the food sector, HSI is used for quality control purposes, including the identification of defects or contaminants in food products (Abbasi et al., 2025; Kheiralipour et al., 2014; Ordoudi et al., 2014; Sunil et al., 2023; Wu et al., 2023). Analysis using advanced machine learning models demonstrate the cross-sector potential of these techniques, which are equally applicable to food quality assessment (Xu et al., 2025; Nargesi and Kheiralipour, 2025). Altogether, these capabilities highlight HSI as a versatile and valuable tool across diverse agricultural and food-related applications (Esmaeily et al., 2024; Khan et al., 2022; Lu et al., 2020; Mahesh et al., 2015; Wang et al., 2023). The continuous evolution of deep learning architectures for spectral image analysis and the detection of small targets in complex spectral data (Li, Liu, Cheng, et al., 2026; Li, Liu, Huang, et al., 2026) demonstrates the growing sophistication of techniques for extracting subtle features from hyperspectral imagery.

The aim of this research is to evaluate the effectiveness of HIS systems in developing predictive models for the rapid and non-destructive estimation of chemical properties of date syrup, which are key indicators of its quality. This technology enables product authentication in terms of ingredient composition. Although several studies have investigated the physicochemical properties of date syrup using conventional analytical methods, comprehensive, rapid, and non-destructive approaches for quality assessment are still limited. To the best of our knowledge, this study represents one of the first systematic investigations employing HIS to characterize the quality attributes of date syrup. HIS, by enabling precise analysis of date syrup, holds great potential for the development of portable and on-site systems that can serve as effective tools for quality control and determination of chemical parameters in the food industry.

2. Materials and methods

The research steps from preparing the test samples to analyzing the data are shown in Fig. 1.

Fig. 1.

Fig. 1

Open in a new tab

Research process steps.

2.1. Sample preparation and handling chemical tests

Three different grades of date syrup under the brand name Mardasang were purchased from a local market in Bandar Abbas, Hormozgan, Iran. Chemical analyses confirmed that these samples represented distinct grades, as they differed in key parameters including acidity, Brix, fructose content, hydroxymethylfurfural (HMF), pH, and viscosity. The collected samples were stored at room temperature until analysis (Fig. 1).

In this study, the properties of acidity, Brix, fructose, Hydroxymethylfurfural (HMF), pH, and viscosity of date syrup were measured using standard laboratory methods and specialized instruments at the Central Laboratory No. 2, Lorestan University, Khoramabad, Lorestan, Iran.

All chemical properties of date syrup samples were determined following the Association of Official Analytical Chemists (AOAC, 1997). Acidity was measured in terms of malic acid. The Brix test was conducted to measure the total soluble solids (mainly sugars) in the date syrup. This is one of the most critical quality control tests for sugary products, as it directly relates to the sweetness, concentration, and overall quality of the product. Brix values were measured using a refractometer. The results were used to assess the natural sugar content, detect possible adulteration (such as the addition of sugar or syrups), and ensure compliance with packaging and export quality standards. Fructose, one of the main sugars in date syrup (alongside glucose), is a key indicator in quality control, nutritional evaluation, and fraud detection. In this study, fructose content was measured using an enzymatic method. HMF is a compound formed due to excessive heating or prolonged storage of sugary products. Elevated HMF levels may indicate poor processing, improper storage, or thermal adulteration. Measurement of pH is important for assessing the microbial and chemical stability of the product. The pH level of the date syrup was measured using a pH meter (Model: pH 7110 model, WTW company inolab) at 20 °C.

Viscosity is a key physical property of date syrup, directly related to its concentration, Brix value, and sensory characteristics. In this study, viscosity was measured using a Brookfield viscometer, in accordance with international standards such as ASTM D2196 or ISO 2555. Higher viscosity indicates better consistency and product quality, making it an important factor in the formulation of food products.

2.2. Hyperspectral scanning

Hyperspectral images of the date syrup samples were captured using a hyperspectral camera (Model: Specam, Parto Sanat Company, Zanjan, Iran) in the wavelength range of 400–950 nm. As shown in Fig. 2, each sample was placed on a motorized platform, and its hyperspectral image was captured individually. This process was carried out through linear scanning at a constant speed using four 25-watt LED lamps. A white reference plate was used for the white reference, and a dark reference was automatically applied during image capturing when the lamps were turned on. The raw hyperspectral images (with 666 bands) were calibrated using the white and dark references before data analysis.

Fig. 2.

Fig. 2

Open in a new tab

Hyperspectral imaging camera.

2.3. Processing of hyperspectral data

Hyperspectral images were processed using MATLAB software version 2016. A region of interest with dimensions 50 × 80 pixels from the center of the raw hyperspectral image was selected to isolate the data corresponding from the container's boundaries to the samples and extract spectral information.

To reduce data complexity, improve analysis accuracy, and minimize data volume, the selection of relevant features was essential. For this purpose, Principal Component Analysis (PCA) was employed, which reduces the resolution by averaging pixel values within each block. The selection of effective channels was based on the average of the first (PC1) and second (PC2) principal components, with key image channels being determined from the peak points in the PC1 and PC2 curves (Fernandez et al., 2019; Kheiralipour et al., 2021; Hasan & Mohieldein, 2019; Nargesi et al., 2026). Then the corresponding wavelengths were found. After selecting appropriate wavelengths, various features were extracted from the selected channels including mean, minimum, maximum, median, variance, and standard deviation (Kheiralipour et al., 2014; Nargesi, 2024). From each hyperspectral image, 7 effective wavelengths were chosen, and 6 features (Mean, Minimum, Maximum, Median, Variance, and Standard deviation) were extracted from each wavelength. In total, 42 effective features were extracted from the hyperspectral images. In cases of high-dimensional data, dimensionality reduction techniques were employed to enhance model performance. In this study, efficient features were selected from the extracted features using sequential feature selection (SFS) method (Nargesi, Kheiralipour, & Jayas, 2024). This method operates based on the model fit criterion, which is the generalized residual sum of squares (Lee et al., 2025; Nargesi & Kheiralipour, 2024).

2.4. Model development for predicting the syrup's properties

In this study, HIS data were used as input features, and laboratory-measured chemical pproperties of date syrup including acidity, Brix, fructose, hydroxymethylfurfural (HMF), pH, and viscosity were used as labeled output variables. Three predictive algorithms were designed and trained: artificial neural network (ANN), partial least squares regression (PLSR), and support vector regression (SVR). Model performance was evaluated based on the accuracy of their predictions. In the ANN model, the activation functions used for the hidden and output layers were tansig and purelin, respectively. The Levenberg-Marquardt training algorithm was employed due to its high convergence speed and its ability to reduce the risk of overfitting (Nargesi, Amiriparian, et al., 2024; Nargesi, Heidarbeigi, et al., 2024). For the SVR model, the hyperparameters C, lambda, epsilon, and kernel width were initially chosen based on commonly recommended values in the literature and then fine-tuned through a grid search to optimize prediction accuracy. The final values were set to C = 100, lambda = 1, epsilon = 0.001, and Gaussian kernel with a width of 0.01, as these provided the best balance between bias and variance. For the PLSR and SVR models, 80% of the data were allocated for training and 20% for testing. In contrast, for the ANN model, the dataset was randomly divided into three subsets: 60% for training, 20% for validation, and 20% for testing. This difference was due to the need for an independent validation set in ANN training to monitor convergence and prevent overfitting, while PLSR and SVR do not require a separate validation set under the cross-validation framework employed (Nargesi & Kheiralipour, 2024; Nargesi, Kheiralipour, & Jayas, 2024).

3. Results and discussion

3.1. Optimal hypercubes' channels

Fig. 3 presents the PC1 and PC2 plots of the hyperspectral image channels (comprising 665 channels) for date syrup samples. As shown in Fig. 3, the values of the principal components exhibit significant variations across the different grades of date syrup. Based on the prominent peaks observed in these plots, the effective channels were identified and selected. The corresponding optimal wavelengths for these channels were then found in the HIS system. The selected effective channels for the different grades of date syrup, as shown in Fig. 3, were 434, 532, 571, 592, 598, 645, and 654. Furthermore, the effective wavelengths corresponding to the selected channels for the three grades of date syrup were identified as 758.43, 839.46, 871.70, 889.07, 894.03, 932.89, and 940.33 nm. In the next step, six features were extracted from the effective image channels. By selecting seven effective wavelengths, 42 features were extracted from the hyperspectral images. Only these features associated with the selected effective wavelengths-identified as having the highest relevance to the classification task-were used to reduce model complexity and enhance accuracy. The features were used for developing predictive models considering two strategies, with and without feature selection step.

Fig. 3.

Fig. 3

Open in a new tab

The first (PC1) and second (PC2) principal components graphs for date syrup.

3.2. Selection of efficient features

From each effective hyperspectral image channel, six statistical features were extracted to provide a more detailed representation of the spectral information. By selecting seven wavelengths as the most influential (effective wavelengths), 42 features were obtained. These extracted features, which served as input of predictive models, are presented in Table 1. These features were used as input of predictive models considering with and without feature selecting step. Feature selection was done to find efficient features and achieve high accuracy and low complicity of the predictive models.

Table 1.

The selected features used to estimate chemical properties of date syrup.

Efficient feature Effective wavelength (nm)
Maximum 722.05
Maximum 811.35
Variance 811.35
Median 839.46
Mean 882.46
Mean 889.07
Mean 901.47
Mean 932.89
Open in a new tab

3.3. Prediction

3.3.1. Chemical properties

The chemical properties of three different date syrup samples were determined through laboratory analyses. The results of these measurements are presented in Table 2. The obtained values were used as output (dependent) variables for training and evaluating the predictive models.

Table 2.

Chemical properties of date syrup.

Syrup grade Acidity (mg/100 g) BRIX (Degrees) Fructose (%) HMF (mg/Kg) PH (−) Viscosity (cm2/s)
1 0.249 69.1 51.8 15.36 7.36 55.3
2 0.345 72.7 53.0 13.56 6.32 60.3
3 0.589 79.2 51.5 12.06 5.84 66.8
Open in a new tab

Prediction of chemical properties of date syrup based on hyperspectral images was carried out using two different strategies: with and without feature selection step. In the strategy of without feature selection step, all features extracted from the hyperspectral images in the effective channels were used as input for the predictive models.

3.3.2. Artificial neural network models

  • 1.

    Based on feature selection

In the strategy with feature selection step, only the efficient features were utilized for prediction. The prediction results for the properties of acidity, Brix, fructose, HMF, pH, and viscosity of date syrup are presented in Fig. 4. In this model, the number of neurons in the input layer of the network was optimized to match the number of effective features and the number of neurons in the output layer corresponded to the number of predicted parameters. Additionally, the number of neurons in the hidden layer varied between 2 and 20 neurons. Various network structures were examined and evaluated to determine the optimal architecture.

Fig. 4.

Fig. 4

Fig. 4

Open in a new tab

Regression plots showing the relationship between predicted and measured values obtained by the optimized artificial neural networks model based on the selected features for quality parameters including: (a) acidity, (b) Brix, (c) fructose, (d) hydroxymethylfurfural, (e) pH, and (f) viscosity.

The performance of the network in the validation phase for the chemical properties of acidity, Brix, fructose, HMF, pH, and viscosity of date syrup in different numbers of iterations is shown in Table 3. It can be seen that after running the network, the least validation error occurred.

Table 3.

The best performance of the artificial neural network based on the selected features.

Acidity Brix Fructose HMF pH Viscosity
Neurons in hidden layer 13 100 7 7 15 5
Network performance 1.25 × 10−14 2.98 × 10−4 9.22 × 10−6 3 × 10−4 4.05 × 10−12 1.01 × 10−2
Open in a new tab

The prediction results for the chemical properties —acidity, Brix, fructose, HMF, pH, and viscosity—in date syrup are presented in Fig. 4. In this section, the optimal architecture of the ANN was designed and implemented for predicting these chemical properties. The correlation coefficients between the predicted and actual values across the training, validation, testing, and overall phases are reported in Table 4 and Fig. 4, demonstrating the high predictive performance of the model for these parameters.

  • 2.

    Without feature selection

Table 4.

The correlation results of the optimal artificial neural netwirks model based on the selected features to predict sucrose of date syrup.

Training Validation Test Total
Acidity 100.00 100.00 100.00 100.00
Brix 100.00 99.99 99.99 99.99
Fructose 100.00 99.99 100.00 100.00
HMF 100.00 99.99 99.99 99.99
pH 100.00 100.00 100.00 100.00
Viscosity 100.00 99.99 99.99 99.99
Open in a new tab

Prediction of hyperspectral images was performed with feature selection strategy. The prediction results for the properties of acidity, Brix, fructose, HMF, pH, and viscosity in date syrup are presented in Fig. 5. Different network structures were evaluated to determine the optimal architecture. The network's performance during the validation phase for the chemical properties of acidity, Brix, fructose, HMF, pH, and viscosity of date syrup using the effective feature selection method at various epochs is shown in Table 5. It can be observed that after running the network, the least validation error occurred.

Fig. 5.

Fig. 5

Fig. 5

Open in a new tab

Regression plots showing the relationship between predicted and measured values obtained by the optimized artificial neural networks model based on all extracted features for quality parameters including, (a) acidity, (b) Brix, (c) fructose, (d) hydroxymethylfurfural, (e) pH, and (f) viscosity.

Table 5.

The best performance of the artificial neural network based on all extracted features.

Acidity Brix Fructose HMF pH Viscosity
Neurons in hidden layer 56 67 16 15 91 10
Network performance 3.8 × 10−14 1.1 × 10−14 3.0 × 10−15 4.9 × 10−15 6.3 × 10−9 6.4 × 10−7
Open in a new tab

The prediction outcomes for the chemical properties —acidity, Brix, fructose, HMF, pH, and viscosity of date syrup are illustrated in Fig. 5. The correlation coefficients between the predicted and actual values across the training, validation, testing, and overall phases are presented in Table 6 and Fig. 5, highlighting the strong predictive performance of the model for these parameters.

Table 6.

The correlation results of the optimal artificial neural networks model based on all extracted features to predict sucrose of date syrup.

Training Validation Test Total
Acidity 100.00 100.00 100.00 100.00
Brix 100.00 100.00 100.00 100.00
Fructose 100.00 100.00 100.00 100.00
HMF 100.00 100.00 100.00 100.00
pH 100.00 100.00 100.00 100.00
Viscosity 100.00 100.00 100.00 100.00
Open in a new tab

3.3.3. Partial least squares regression

  • 1.

    Without feature selection

In this study, PLSR was employed to analyze the chemical properties of date syrup using all extracted features from the effective wavelength regions. The data were obtained from two Excel files containing the spectral features of date syrup and the corresponding chemical parameter values. After implementing the PLSR model, the principal input and output components were identified, and metrics such as the explained variance percentage and Mean Squared Error (MSE) were calculated. The correlation coefficient between the predicted and actual values for properties including acidity, Brix, fructose, HMF, pH, and viscosity was 99.78, 60.60, 11.76, 98.59, 22.40, and 93.66%, with an MSE of 5.3 × 10−5, 65.90, 12.12, 0.07, 0.99, and 21.35 indicating the model's high accuracy in predicting the dependent variables (Fig. 6).

Fig. 6.

Fig. 6

Open in a new tab

Correlation results of the partial least squares regression model based on all extracted features.

To predict the chemical properties extracted from hyperspectral images of date syrup, the PLSR model was employed. In this process, the principal input and output components were optimally selected to reduce data dimensionality while preserving relevant information. Model performance was evaluated using statistical metrics such as explained variance percentage and Mean Squared Error (MSE). Parameter prediction was based on regression coefficients (β), and the model's accuracy was assessed through MSE. The correlation coefficient (r) indicated the agreement between predicted and actual values, while the coefficient of determination (R2) quantified the model's explanatory power as a percentage of the variance in the dependent variables.

  • 2.

    Based on feature selection

In this study, PLSR was applied to model and predict six chemical properties of date syrup, including acidity, Brix, fructose, HMF, pH, and viscosity. The results demonstrated that the model achieved a coefficient of determination of 99.63, 99.65, 99.2, 99.66, 99.64, and 99.66% and an MSE of 7.5 × 10−5, 0.06, 0.003, 0.006, 0.001, 0.08 indicating high predictive accuracy and confirming the effectiveness and reliability of the proposed approach (Fig. 7).

Fig. 7.

Fig. 7

Open in a new tab

Correlation results of the partial least squares regression model based on the selected features.

3.3.4. Support vector regression models

  • 1.

    Without feature selection

In this study, SVR was employed to predict the chemical properties of date syrup, including acidity, Brix, fructose, HMF, pH, and viscosity, using all relevant spectral features. The model was configured with hyperparameters including C = 100C = 100C = 100, λ = 1\lambda = 1λ = 1, ϵ = 0.001\epsilon = 0.001ϵ = 0.001, a kernel width of 0.01, and a Gaussian kernel function. The dataset was randomly divided into training and testing subsets to evaluate model performance.

The predictive performance of the SVR model was assessed using statistical metrics such as Mean Squared Error (MSE) and the coefficient of determination (R2). The model achieved an MSE of 0.02, 1.6 × 10−3, 0.01, 0.2, 0.06, and 1.02 × 10−3 and an R2 of 95.47, 88.90, 99.78, 98.70, 97.67, and 82.77%, indicating a reasonably high level of accuracy in estimating the chemical properties of date syrup based on hyperspectral data (Fig. 8).

  • 2.

    Based on feature selection

Fig. 8.

Fig. 8

Fig. 8

Open in a new tab

Correlation results of the support vector regression model based on all extracted features.

In this study, SVR was employed to quantitatively analyze the chemical compositions of date syrup, including acidity, Brix, fructose, HMF, pH, and viscosity, based on effective features extracted from the spectral data. To evaluate the model's performance, the coefficient of determination (R2) was calculated by comparing the predicted and actual values of the chemical properties. The R2 values for the respective compounds were 98.93, 97.93, 36.16, 93.46, 93.44, and 91.10%. Additionally, the model's error was assessed using the Mean Squared Error (MSE) metric, with MSE values of 0.05, 2.1, 189.86, 0.06, 0.04 and 4.2 for the corresponding parameters (Fig. 9).

Fig. 9.

Fig. 9

Fig. 9

Open in a new tab

Correlation results of the support vector regression model based on the selected features.

Artificial neural network and PLSR models were developed to predict Brix, acidity, and the Brix-to-acidity ratio in mango fruit. These models achieved a coefficient of determination (R2) ranging from 0.50 to 0.89, indicating acceptable prediction accuracy. Additionally, PLSR models yielded R2 values equal to or greater than 0.60 in many cases. HSI offers a rapid and non-destructive method for predicting the internal quality of lychee and mango flesh. This approach can assist in the timely scheduling of harvest and enable precise grading of fruits into uniform-quality batches (Kanwal et al., 2025).

Conventional methods for determining ash content in coal, based on proximate analysis protocols, are time-consuming, costly, and require destructive procedures such as direct combustion of samples. This study introduces an innovative and non-destructive approach utilizing mid-infrared Fourier Transform Infrared (FTIR) spectroscopy in the range of 1450 to 350 cm−1 to identify spectrally sensitive regions (comprising fourteen selective absorption bands) for accurate prediction of ash yield in coal samples. A set of advanced machines learning algorithms, including Piecewise Linear Regression (PLR), Artificial Neural Networks (ANN), PLSR, SVR, and Random Forest (RF), were employed for predictive modeling. Additionally, a multi-model estimation (MME) framework was proposed, combining the outputs of the top three performing models (PLR, PLSR, and ANN) to enhance prediction accuracy and model robustness. The resulting hybrid model demonstrated superior performance, achieving a coefficient of determination (R2) of 0.883, root mean square error (RMSE) of 3.059 wt%, RMSE percentage (RMSE%) of 30.080, mean bias error percentage (MBE%) of 3.694, and mean absolute error (MAE) of 2.249 wt%. These findings highlight the high potential of FTIR spectroscopy as a rapid and reliable tool for non-destructive ash yield prediction in coal, underscoring its applicability as an efficient analytical solution for quality monitoring in industrial coal processing (Mishra et al., 2025).

In the analysis of hyperspectral data for predicting the chemical properties of date syrup, the performance of three machine learning models, namely Artificial Neural Network (ANN), PLSR, and SVR, was compared. The results indicated that ANN consistently outperformed the other two models across all chemical parameters, achieving prediction accuracies close to or at 100%. In contrast, PLSR and SVR provided acceptable predictions but exhibited slightly lower accuracies, particularly for parameters with complex nonlinear relationships with the hyperspectral data, such as viscosity and HMF content.

The superior performance of ANN can be attributed to several factors. First, hyperspectral data often exhibit complex nonlinear relationships with chemical properties, and ANN is inherently capable of modeling these nonlinearities through multiple layers and nonlinear activation functions, whereas PLSR assumes linear relationships and SVR requires careful kernel selection. Second, ANN can efficiently process high-dimensional data without extensive dimensionality reduction, while PLSR relies on latent variable extraction and SVR may suffer from overfitting in the presence of irrelevant features. Third, many hyperspectral bands are highly correlated; ANN, through its network structure and learning algorithms, can handle multicollinearity more effectively, whereas PLSR is more sensitive to highly correlated predictors.

Although the inclusion of a feature selection step slightly improved the performance of PLSR and SVR by reducing input dimensionality and noise, ANN achieved near-optimal results even without this step. This highlights the model's intrinsic ability to directly extract the most informative spectral patterns from the full hyperspectral dataset. Overall, these results indicate that ANN is particularly well-suited for hyperspectral applications in food analysis, providing high prediction accuracy, robustness to complex nonlinearities, and adaptability to high-dimensional spectral data.

4. Conclusions

In this study, hyperspectral imaging was employed to estimate six chemical properties of date syrup, including acidity, Brix, fructose, HMF, pH, and viscosity. Three prediction algorithms Artificial Neural Networks, Partial Least Squares Regression, and Support Vector Regression were implemented and evaluated using two strategies: with and without feature selection. Initially, the effective spectral channels were identified, and six prominent features were extracted from each of the seven selected wavelengths, yielding 42 initial features. Feature selection reduced this set to eight key features, significantly simplifying the models while maintaining predictive power. The results demonstrated that Artificial Neural Networks models using the selected features outperformed other algorithms, achieving correlation coefficients (R2) of 99.99–100% across all properties and the lowest validation errors. Partial Least Squares Regression and Support Vector Regression also showed strong predictive performance, particularly when feature selection was applied, with R2 values exceeding 91% for most properties. Quantifying the practical benefits, the Artificial Neural Networks based system reduced the number of input features by 81% (from 42 to 8), which can substantially decrease computation time during real-time monitoring. The high prediction accuracy and reduced model complexity indicate that such systems could shorten detection time by approximately 30–40% while maintaining near-perfect prediction of chemical properties. These results highlight the potential of hyperspectral imaging combined with ANN for developing intelligent, real-time quality control systems in the food processing industry, capable of both optimizing operational efficiency and enhancing the reliability of chemical property monitoring in date syrup.

CRediT authorship contribution statement

Mohammad Hossein Nargesi: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Kamran Kheiralipour: Writing – review & editing, Writing – original draft, Software, Methodology, Formal analysis.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

Acknowledgment

The authors thank Mechanical Engineering of Biosystems Department, Ilam University, Ilam, Iran, for their supports in the present research.

Ethical statement

No studies involving humans or animals were conducted in the present research.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Contributor Information

Mohammad Hossein Nargesi, Email: mohamadhossein.73@gmail.com.

Kamran Kheiralipour, Email: k.kheiralipour@ilam.ac.ir.

Data availability

Data will be made available on request.

References

  1. Abbasi K., M.H., Karami P., Kheiralipour K. Differentiating Alternaria species using hyperspectral imaging. Biomechanism and Bioenergy Research. 2025;4(3):1–10. [Google Scholar]
  2. Ahmed T., Wijewardane N.K., Lu Y., et al. Advancing sweetpotato quality assessment with hyperspectral imaging and explainable artificial intelligence. Computers and Electronics in Agriculture. 2024;220 doi: 10.1016/j.compag.2024.108855. [DOI] [Google Scholar]
  3. Al-Farsi K., Al-Habsi N.A., Al-Khusaibi M. The potential antioxidant properties of date products: A concise update. Canadian Journal of Clinical Nutrition. 2018;6:84–104. [Google Scholar]
  4. Al-Farsi M., Alasavar C., Al-Shoaily K., Al-Amry M., Al-Rawahy F. Compositional and functional characteristics of dates, syrup, and there by – Products. Food Chemistry. 2007;104:943–947. [Google Scholar]
  5. Al Hilfi M.K., Al-Fekaiki D.F., Al-Hilphy A.R. Identification and determination of metal elements of dates syrup extracted from various varieties using SEM-EDS technique. Basrah Journal of Agricultural Sciences. 2019;32(2):126–134. [Google Scholar]
  6. Alhamdan A.M., Al Juhaimi F.Y., Hassan B.H., Ehmed K.A., Mohamed Ahmed I.A. Physicochemical, microbiological, and sensorial quality attributes of a fermented milk drink (Laban) fortified with date syrup (dibs) during cold storage. Foods. 2021;10(12):3157. doi: 10.3390/foods10123157. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Alkhoori M.A., Kong A.S.Y., Aljaafari M.N., Abushelaibi A., Erin Lim S.H., Cheng W.H., et al. Biochemical composition and biological activities of date palm (Phoenix dactylifera L.) seeds: A review. Biomolecules. 2022;12(11) doi: 10.3390/biom12111626. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. AOAC . 16th ed. Association of Official Analytical Chemists; Washington, DC: 1997. Official methods of analysis of AOAC international. [Google Scholar]
  9. Ashraf Z., Hamidi-Esfahani Z. Date and date processing: A review. Food Review International. 2011;27(2):101–133. [Google Scholar]
  10. Barreveld W.H. 1993. Date palm products. Numéro Agricole 101 de Bulletin de Services de la FAO. [Google Scholar]
  11. Caliskan O., Polat A.A. Phytochemical and antioxidant properties of selected fig (Ficus carica L.) accessions from the eastern Mediterranean region of Turkey. Scientia Horticulturae. 2011;128:473–478. [Google Scholar]
  12. Cepeda E., Villaran M.C. Density and viscosity of malus Floribunda juice as a function of concentration and emperature. Journal of Food Engineering. 1999;41:103–107. [Google Scholar]
  13. Deng B., He Y., Shen Z., Zhang Y., Deng Q., Nie Z., Wang Y. YCNNet: Road target recognition method by fusion of LiDAR and thermal infrared camera. IEEE Sensors Journal. 2026;26(2):3278–3288. doi: 10.1109/JSEN.2025.3639441. [DOI] [Google Scholar]
  14. El-Beltagy A.E., Nassar A.G., El-Ghobashy A.K., Yousef H.Y.M. Microwave a potent date syrup producing method. Egyptian Journal of Applied Science. 2009;24(8B):454–464. [Google Scholar]
  15. Elmasry G., Kamruzzaman M., Sun D.W., Allen P. Principles and applications of hyperspectral imaging in quality evaluation of Agro-food products: A review. Critical Reviews in Food Science and Nutrition. 2012;52:999–1023. doi: 10.1080/10408398.2010.543495. [DOI] [PubMed] [Google Scholar]
  16. El-Nagga E., Abd El–Tawab Y. Compositional characteristics of date syrup extracted by different methods in some fermented dairy products. Annals of Agricultural Sciences. 2012;57(1):29–36. [Google Scholar]
  17. Esmaeily R., Razavi M.A., Razavi S.H. A step forward in food science, technology and industry using artificial intelligence. Trends in Food Science and Technology. 2024;143 doi: 10.1016/j.tifs.2023.104286. [DOI] [Google Scholar]
  18. Farahnaky A., Mardani m., Mesbahi G., Majzoobi M., Golmakani M.T. Some physicochemical properties of date syrup, concentrate, and liquid sugar in comparison with sucrose solutions. Journal of Agricultural Science and Technology. 2016;18:657–668. [Google Scholar]
  19. Fernandez L.C., Allende-Prieto J., Peon E. Preliminary assessment of visible, near infrared, and short-wavelength–infrared spectroscopy with a portable instrument for the detection of Staphylococcus aureus biofilms on surfaces. Journal of Food Protection. 2019;82(8):1314–1319. doi: 10.4315/0362-028X.JFP-18-567. [DOI] [PubMed] [Google Scholar]
  20. Gad A.S., Kholif A.M., Sayed A.F. Evaluation of the nutritional value of functional yogurt resulting from combination of date palm syrup and skim milk. American Journal of Food Technology. 2010;5(4):250–259. [Google Scholar]
  21. Gholamhosseinpour A., Hashemi S.M.B. Optimization of fermentation process of date syrup by lactobacillus delbrueckii and lactobacillus acidophilus: Microbial growth, carbohydrate metabolism, and peptide content. Journal of Food Processing and Preservation. 2022;46(4) [Google Scholar]
  22. Hasan M., Mohieldein A. In vivo evaluation of anti-diabetic, hypolipidemic, antioxidative activities of Saudi date seed extract on streptozotocin induced diabetic rats. Journal of Clinical and Diagnostic Research: JCDR. 2019;10(3):FF06. doi: 10.7860/JCDR/2016/16879.7419. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. He H.J., Sun D.W. Hyperspectral imaging technology for rapid detection of various microbial contaminants in agricultural and food products. Trends in Food Science and Technology. 2015;46:99–109. doi: 10.1016/j.tifs.2015.08.001. [DOI] [Google Scholar]
  24. Ishurd O., Kennedy J.F. The anti-cancer activity of polysaccharide prepared from Libyan dates (Phoenix dactylifera L.) Carbohydrate Polymers. 2005;59(4):531–535. [Google Scholar]
  25. Jia J., Zheng X., Wang Y., Chen Y., Karjalainen M., Dong S.…Hyyppä J. The effect of artificial intelligence evolving on hyperspectral imagery with different signal-to-noise ratio, spectral and spatial resolutions. Remote Sensing of Environment. 2024;311 doi: 10.1016/j.rse.2024.114291. [DOI] [Google Scholar]
  26. Kamruzzaman M. Optical sensing as analytical tools for meat tenderness measurements - A review. Meat Science. 2023;195 doi: 10.1016/j.meatsci.2022.109007. [DOI] [PubMed] [Google Scholar]
  27. Kamruzzaman M., Makino Y., Oshita S. Non-invasive analytical technology for the detection of contamination, adulteration, and authenticity of meat, poultry, and fish: A review. Analytica Chimica Acta. 2015;853:19–29. doi: 10.1016/j.aca.2014.08.043. [DOI] [PubMed] [Google Scholar]
  28. Kanwal N., Kämper Y., Farrar M.B., Tootoonchy M., Lynch C., Nichols J.…Hosseini Bai S. Rapid assessment of lychee and mango fruit quality using hyperspectral imaging. LWT. 2025;224 doi: 10.1016/j.lwt.2025.117833. [DOI] [Google Scholar]
  29. Keshtkaran M., Mohammadifar M.A. The effects of date syrup and gum tragacanth on physical and rheological properties of date milk beverage. World Journal of Dairy & Food Sciences. 2013;8(2):147–155. doi: 10.3168/jds.2012-5942. [DOI] [PubMed] [Google Scholar]
  30. Khan A., Vibhute A.D., Mali S., Patil C.H. A systematic review on hyperspectral imaging technology with a machine and deep learning methodology for agricultural applications. Ecological Informatics. 2022;69 doi: 10.1016/j.ecoinf.2022.101678. [DOI] [Google Scholar]
  31. Kheiralipour, K. (2012). Implementation and construction of a system for detecting fungal infection of pistachio kernel based on thermal imaging (TI) and image processing technology. Ph.D. Dissertation, University of Tehran, Karaj, Iran.
  32. Kheiralipour K. Sustainable production: Definitions, aspects, and elements. 1st ed. Nova Science Publishers; Hauppauge, New York, USA: 2022. [Google Scholar]
  33. Kheiralipour K. Nova Science Publishers; Hauppauge, New York, USA: 2024. The future of imaging technology. ISBN 979-8-89530-078-7. [Google Scholar]
  34. Kheiralipour K., Ahmadi H., Rajabipour A., Rafiee S., Javan-Nikkhah M., Jayas D.S. 8th Iranian National Congress of agricultural machinery engineering (biosystems) and mechanization. 29–31 January, Mashahd, Iran. 2014. Detection of healthy and fungal-infected pistachios based on hyperspectral image processing. [Google Scholar]
  35. Kheiralipour K., Ahmadi H., Rajabipour A., Rafiee S., Javan-Nikkhah M., Jayas D.S., Siliveru K., Malihipour A. Processing the hyperspectral images for detecting infection of pistachio kernel by R5 and KK11 isolates of aspergillus flavus fungus. Iranian Journal of Biosystems Engineering. 2021;52(1):13–25. [Google Scholar]
  36. Kheiralipour K., Chelladurai V., Jayas D.S. In: Image processing: Advances in applications and research. Jayas D.S., editor. Nova Science Publishers; Hauppauge, New York, USA: 2023. Imaging systems and image processing techniques. (ISBN: 979-8886979756) [Google Scholar]
  37. Kheiralipour K., Jayas D.S. Current and future applications of hyperspectral imaging in agriculture, nature and food. Trends in Technical & Scientific Research. 2024;7(2):1–9. [Google Scholar]
  38. Kheiralipour K., Sajadipour F., Nadimi M. A review of nut quality assessment using hyperspectral imaging technique. Journal of Food Composition and Analysis. 2025;148(1) [Google Scholar]
  39. Kheiralipour K., Sajadipour F., Nargesi M.H. Applications of spectral imaging in biosystems engineering in Iran, a review. Recent Progress in Sciences. 2025;2(1) [Google Scholar]
  40. Kheiralipour K., Singh C.B., Jayas D.S. In: Image processing: Advances in applications and research. Jayas D.S., editor. Nova Science Publishers; Hauppauge, New York, USA: 2023. Applications of visible, thermal, and hyperspectral imaging techniques in the assessment of fruits and vegetables. (ISBN: 979-8886979756) [Google Scholar]
  41. Lee C.M., Lin Y.F., Ho Y.H., Kang L.W., Hsu C.C. HyFusion: Enhanced reception field transformer for hyperspectral image fusion. Computer Vision and Pattern Recognition. 2025 doi: 10.48550/arXiv.2501.04665. [DOI] [Google Scholar]
  42. Lee G.Y., Dam T., Ferdaus M.M., et al. Unlocking the capabilities of explainable few-shot learning in remote sensing. Artificial Intelligence Review. 2024;57 doi: 10.1007/s10462-024-10803-5. [DOI] [Google Scholar]
  43. Li L., Liu L., Cheng F., He Y., Zhong Z. CN-UNet: ConvNeXt UNet with slicing-aided hyper segmentation for infrared small target detection. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. 2026;19:84–98. doi: 10.1109/JSTARS.2025.3631384. [DOI] [Google Scholar]
  44. Li L., Liu L., Huang D., Wang S., Wang X., He Y., Zhong Z. KECS-net: Knowledge-embedded CSwin-UNet with slicing-aided hypersegmentation for infrared small target detection. IEEE Geoscience and Remote Sensing Letters. 2026;23:1–5. doi: 10.1109/LGRS.2025.3632827. [DOI] [Google Scholar]
  45. Lu B., Dao P., Liu J., He Y., Shang J. Recent advances of hyperspectral imaging technology and applications in agriculture. Remote Sensing. 2020;12:2659. doi: 10.3390/rs12162659. [DOI] [Google Scholar]
  46. Mahesh S., Jayas D.S., Paliwal J., White N.D.G. Hyperspectral imaging to classify and monitor quality of agricultural materials. Journal of Stored Products Research. 2015;61:17–26. doi: 10.1016/j.jspr.2015.01.006. [DOI] [Google Scholar]
  47. Maqsood S., Adiamo O., Ahmad M., Mudgil P. Bioactive compounds from date fruit and seed as potential nutraceutical and functional food ingredients. Food Chemistry. 2020;308(5) doi: 10.1016/j.foodchem.2019.125522. [DOI] [PubMed] [Google Scholar]
  48. Mishra S., Prasad A.K., Vinod A., Shukla A., Mukherjee S., Purkait B.…Sarkar B.C. A multi-model approach for estimation of ash yield in coal using Fourier transform infrared spectroscopy. Scientific Reports. 2025;15:13786. doi: 10.1038/s41598-025-98071-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Nargesi M.H. 2024. Detection of fraud in black pepper, red pepper, and cinnamon powder using hyperspectral imaging and artificial neural network. Ph.D. Dissertation, University of Bu Ali Sina, Hamadan, Iran. [Google Scholar]
  50. Nargesi M.H., Amiri Parian j., Kheiralipour K. Detection of different adulteration levels in red pepper using hyperspectral imaging combined with machine learning algorithms. Sensing and Bio-Sensing Research. 2026;52:101016. [Google Scholar]
  51. Nargesi M.H., Amiriparian J., Bagherpour H., Kheiralipour K. Detection of different adulteration in cinnamon powder using hyperspectral imaging and artificial neural network method. Results in Chemistry. 2024;9 [Google Scholar]
  52. Nargesi M.H., Aziznia S., Kheiralipour K. Hyperspectral imaging feasibility in the prediction of melanoidin parameters in whey protein-β-glucan conjugates. LWT. 2026;250:119460. [Google Scholar]
  53. Nargesi M.H., Heidarbeigi K., Moradi Z., Abdolahi S. Detection of chlorine in potassium chloride and potassium sulfate using chemical imaging and artificial neural network. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy. 2024;326 doi: 10.1016/j.saa.2024.125253. [DOI] [PubMed] [Google Scholar]
  54. Nargesi M.H., Kheiralipour K. Visible feature engineering to detect adulteration in black and red peppers. Scientific Reports. 2024;14 doi: 10.1038/s41598-024-76617-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Nargesi M.H., Kheiralipour K. Estimating ash content in wheat flour using visible-near infrared hyperspectral imaging and machine learning methods. LWT. 2025;234 [Google Scholar]
  56. Nargesi M.H., Kheiralipour K., Jayas D.S. Classification of different wheat flour types using hyperspectral imaging and machine learning techniques. Infrared Physics & Technology. 2024;142 [Google Scholar]
  57. Nargesi M.H., Kheiralipour K., Khodaei J. Intelligent detection of potassium oxide in potassium sulfate fertilizer using hyperspectral imaging and artificial neural networks. Scientific Reports. 2025;15(1):39687. doi: 10.1038/s41598-025-23232-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Ordoudi S.A., Mozos Pascual M., Tsimidou M.Z. On the quality control of traded saffron by means of transmission Fourier-transform mid-infrared (FT-MIR) spectroscopy and chemometrics. Food Chemistry. 2014;150:414–421. doi: 10.1016/j.foodchem.2013.11.014. [DOI] [PubMed] [Google Scholar]
  59. Razavi S.M.A., Najafi M.B.H., Alaee Z. The time independent rheological properties of low-fat sesame paste/date syrup blends as a function of fat substitutes and temperature. Food Hydrocolloids. 2007;21:198–202. [Google Scholar]
  60. Roukas T., Kotzekidou P. Pretreatment of date syrup to increase citric acid production. Enzyme and Microbial Technology. 1997;21:273–276. [Google Scholar]
  61. Shehzad M., Rasheed H., Naqvi S.A., Al-Khayri J.M., Lorenzo J.M., Alaghbari M.A., Manzoor M.F., Aadil R.M. Therapeutic potential of date palm against human infertility: A review. Metabolites. 2021;11 doi: 10.3390/metabo11060408. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Siddeeg A., Zeng X.A., Rahaman A., Manzoor M.F., Ahmed Z., Ammar A.F. Quality characteristics of the processed dates vinegar under influence of ultrasound and pulsed electric field treatments. Journal of Food Science and Technology. 2019;56:4380–4389. doi: 10.1007/s13197-019-03906-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Siddiq M., Aleid S.M., Kader A.A. John Wiley & Sons; 2013. Dates: Postharvest science, processing technology and health benefits. [Google Scholar]
  64. Sun D.W., Pu H., Yu J. Applications of hyperspectral imaging technology in the food industry. Nature Reviews Electrical Engineering. 2024;1:251–263. doi: 10.1038/s44287-024-00033-w. [DOI] [Google Scholar]
  65. Sunil C.K., Jaidhar C.D., Patil N. Systematic study on deep learning-based plant disease detection or classification. Artificial Intelligence Review. 2023;56:14955–15052. doi: 10.1007/s10462-023-10517-0. [DOI] [Google Scholar]
  66. Tu B., Zhou T., Liu B., He Y., Li J., Plaza A. Multi-scale autoencoder suppression strategy for hyperspectral image anomaly detection. IEEE Transactions on Image Processing. 2025;34:5115–5130. doi: 10.1109/TIP.2025.3595408. [DOI] [PubMed] [Google Scholar]
  67. Wang B., Sun J., Xia L., Liu J., Wang Z.H., Li P., Guo Y., Sun X. The applications of hyperspectral imaging technology for agricultural products quality analysis: A review. Food Reviews International. 2023;39:1043–1062. doi: 10.1080/87559129.2021.1929297. [DOI] [Google Scholar]
  68. Wang C., Liu B., Liu L., Zhu Y., Hou J., Liu P., Li X. A review of deep learning used in the hyperspectral image analysis for agriculture. Artificial Intelligence Review. 2021;54:5205–5253. doi: 10.1007/s10462-021-10018-y. [DOI] [Google Scholar]
  69. Wu Q., Mousa M.A.A., Al-Qurashi A.D., Ibrahim O.H.M., Abo-Elyousr K.A.M., Rausch K., Abdel Aal A.M.K., Kamruzzaman M. Global calibration for non-targeted fraud detection in quinoa flour using portable hyperspectral imaging and chemometrics. Current Research in Food Science. 2023;6 doi: 10.1016/j.crfs.2023.100483. [DOI] [PMC free article] [PubMed] [Google Scholar]
  70. Xu H., Xu Y., Hu K. A vision-based inspection system for pharmaceutical production line. The Journal of Supercomputing. 2025;81 doi: 10.1007/s11227-025-07135-8. [DOI] [Google Scholar]
  71. Yahmed N.B., Dauptain K., Lajnef I., Carrere H., Trably E., Smaali E. New sustainable bioconversion concept of date by-products (Phoenix dactylifera L.) to biohydrogen, biogas and date-syrup. International Journal of Hydrogen Energy. 2021;46(1):297–305. [Google Scholar]
  72. Yousif A., Alghamdi A., Hamad A., Mustafa A. Processing and evaluation of a date juice-milk drink. Egyptian Journal of Dairy Science. 1996;24:277–288. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

Data will be made available on request.

Articles from Food Chemistry: X are provided here courtesy of Elsevier

RESOURCES