Prediction of texture characteristics from extrusion food surface images using a computer vision system and artificial neural networks

Formation and prediction of heterocyclic amines and N-nitrosamines in smoked sausages using back propagation artificial neural network

BACKGROUND

Heterocyclic amines (HAs) and N-nitrosamines (NAs) are formed easily during the thermal processing of food, and epidemiological studies have demonstrated that consuming HAs and NAs increases the risk of cancer. However, there are few studies on the application of back propagation artificial neural network (BP-ANN) models to simultaneously predict the content of HAs and NAs in sausages. This study aimed to investigate the effects of cooking time and temperature, smoking time and temperature, and fat-to-lean ratio on the formation of HAs and NAs in smoked sausages, and to predict their total content based on the BP-ANN model.

RESULTS

With an increase in processing time, processing temperature and fat ratio, the content of HAs and NAs in smoked sausages increased significantly, while the content of HA precursors and nitrite residues decreased significantly. The optimal network topology of the BP-ANN model was 5-11-2, the correlation coefficient values for training, validation, testing and all datasets were 0.99228, 0.99785, 0.99520 and 0.99369, respectively, and the mean squared error value of the best validation performance was 0.11326. The bias factor and the accuracy factor were within acceptable limits, and the predicted values approximated the true values, indicating that the model has good predictive performance.

CONCLUSION

The contents of HAs and NAs in smoked sausages were significantly influenced by the cooking conditions, smoking conditions and fat ratio. The BP-ANN model has high application value in predicting the contents of HAs and NAs in sausages, which provides a theoretical basis for the suppression of carcinogen formation. © 2024 Society of Chemical Industry.

Study on the Residence Time and Texture Prediction of Pea Protein Extrusion Based on Image Analysis

This paper initially involves three main processing parameters: screw speed, feeding speed, and initial material moisture content, exploring the RTD of materials inside the extruder barrel under varying parameters and clarifying the impact of parameter variations on RTD. Finally, machine vision technology was utilized to link extruded product images to texture features, and a texture prediction model based on image features was established using a Back Propagation (BP) neural network. Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) were applied to optimize the BP neural network. The results indicate that the feeding speed has a stronger impact than the screw speed on the extrusion process, and an increase in the initial material moisture content tends to shorten the RTD. Specifically, an increase in screw speed results in a denser product structure, while higher feeding speeds lead to reduced pore size in the microstructure. As the initial material moisture content increased from 55% to 70%, the average residence time MRT decreased from 265.21 s to 166.62 s. Additionally, elevated moisture content causes a more porous microstructure. After optimizing the texture prediction model of extruded products through the application of Particle Swarm Optimization and Genetic Algorithm models, it was discovered that the Genetic Algorithm was more effective in reducing errors (p < 0.05) than the Particle Swarm Optimization algorithm. It was found that the Particle Swarm Optimization model exhibited better prediction performance. The results of the prediction indicated a significant association between the image features of the product and hardness, resilience, and chewiness, as corroborated by correlation coefficients of 0.93913, 0.94040, and 0.94724, respectively.

Development of a Low-Cost Artificial Vision System as an Alternative for the Automatic Classification of Persian Lemon: Prototype Test Simulation

In the present research work, an algorithm of artificial neural network (ANN) has been developed based on the processing of digital images of Persian lemons with the aim of optimizing the quality control of the product. For this purpose, the physical properties (weight, thickness of the peel, diameter, length, and color) of 90 lemons selected from the company Esperanza de San José Ornelas SPR de RL (Jalisco, Mexico) were studied, which were divided into three groups (Category "extra", Category I, and Category II) according to their characteristics. The parameters of weight (26.50 ± 3.00 g), diameter/length (0.92 ± 0.08) and thickness of the peel (1.50 ± 0.29 mm) did not present significant differences between groups. On the other hand, the color (determined by the RGB and HSV models) presents statistically significant changes between groups. Due to the above, the proposed ANN correctly classifies 96.60% of the data obtained for each of the groups studied. Once the ANN was trained, its application was tested in an automatic classification process. For this purpose, a prototype based on the operation of a stepper motor was simulated using Simulink from Matlab, which is connected to three ideal switches powered by three variable pulse generators that receive the information from an ANN and provide the corresponding signal for the motor to turn to a specific position. Manual classification is a process that requires expert personnel and is prone to human error. The scientific development presented shows an alternative for the automation of the process using low-cost computational tools as a potential alternative.

Quantitative characterizations of visual fibrousness in meat analogues using automated image analysis

A desirable quality of plant-based meat analogues is to resemble the fibrous structure of cooked muscle meat. While texture analysis can characterize fibrous structures mechanically, assessment of visual fibrous structures remains subjective. Quantitative assessment of visual fibrous structures of meat analogues relies on expert knowledge, is resource-intensive, and time-consuming. In this study, a novel image-based method (Fiberlyzer) is developed to provide automated, quantitative, and standardized assessment of visual fibrousness of meat analogues. The Fiberlyzer method segments fibrous regions from 2D images and extracts fiber shape features to characterize the fibrous structure of meat analogues made from mung bean, soy, and pea protein. The computed fiber scores (the ratio between fiber length and width) demonstrate a strong correlation with expert panel evaluations, particularly on a per-formulation basis (r2  = 0.93). Additionally, the Fiberlyzer method generates fiber shape features including fiber score, fiber area, and the number of fiber branches, facilitating comparisons of structural similarity between meat analogue samples and cooked chicken meat as a benchmark. With a simple measurement setup and user-friendly interface, the Fiberlyzer method can become a standard tool integrated into formulation development, quality control, and production routines of plant-based meat analogue. This method offers rapid, cheap, and standardized quantification of visual fibrousness, minimizing the need for expert knowledge in the process of quality control.

A Reliable Method for Determining the Degree of Orientation of Fibrous Foods Using Laser Transmission and Computer Vision

The degree of organised alignment of fibre structures, referred to as the degree of orientation, significantly influences the textural properties and consumer acceptance of fibrous foods. To develop a new method to quantitatively characterise the fibre structure of such foods, a laser transmission imaging system is constructed to capture the laser beam spot on a sample, and the resulting image undergoes a series of image processing steps that use computer vision to translate the light and dark variations of the original images into distinct ellipses. The results show that the degree of orientation can be reasonably calculated from the ellipse obtained by fitting the outermost isopixel points. To validate the reliability of the newly developed method, we determine the degree of orientation of typical fibrous foods (extruded beef jerky, pork jerky, chicken jerky, and duck jerky). The ranking of the measured orientation agrees with the results of pseudocolour maps and micrographs, confirming the ability of the method to distinguish different fibrous foods. Furthermore, the relatively small coefficients of variation and the strong positive correlation between the degree of organisation and the degree of orientation confirm the reliability of this newly developed method.

Three-Dimensional Printing Parameter Optimization for Salmon Gelatin Gels Using Artificial Neural Networks and Response Surface Methodology: Influence on Physicochemical and Digestibility Properties

This study aimed to optimize the 3D printing parameters of salmon gelatin gels (SGG) using artificial neural networks with the genetic algorithm (ANN-GA) and response surface methodology (RSM). In addition, the influence of the optimal parameters obtained using the two different methodologies was evaluated for the physicochemical and digestibility properties of the printed SGG (PSGG). The ANN-GA had a better fit (R2 = 99.98%) with the experimental conditions of the 3D printing process than the RSM (R2 = 93.99%). The extrusion speed was the most influential parameter according to both methodologies. The optimal values of the printing parameters for the SGG were 0.70 mm for the nozzle diameter, 0.5 mm for the nozzle height, and 24 mm/s for the extrusion speed. Gel thermal properties showed that the optimal 3D printing conditions affected denaturation temperature and enthalpy, improving digestibility from 46.93% (SGG) to 51.52% (PSGG). The secondary gel structures showed that the β-turn structure was the most resistant to enzymatic hydrolysis, while the intermolecular β-sheet was the most labile. This study validated two optimization methodologies to achieve optimal 3D printing parameters of salmon gelatin gels, with improved physicochemical and digestibility properties for use as transporters to incorporate high value nutrients to the body.

Predicting the Textural Properties of Plant-Based Meat Analogs with Machine Learning

Plant-based meat analogs are food products that mimic the appearance, texture, and taste of real meat. The development process requires laborious experimental iterations and expert knowledge to meet consumer expectations. To address these problems, we propose a machine learning (ML)-based framework to predict the textural properties of meat analogs. We introduce the proximate compositions of the raw materials, namely protein, fat, carbohydrate, fibre, ash, and moisture, in percentages and the "targeted moisture contents" of the meat analogs as input features of the ML models, such as Ridge, XGBoost, and MLP, adopting a build-in feature selection mechanism for predicting "Hardness" and "Chewiness". We achieved a mean absolute percentage error (MAPE) of 22.9%, root mean square error (RMSE) of 10.101 for Hardness, MAPE of 14.5%, and RMSE of 6.035 for Chewiness. In addition, carbohydrates, fat and targeted moisture content are found to be the most important factors in determining textural properties. We also investigate multicollinearity among the features, linearity of the designed model, and inconsistent food compositions for validation of the experimental design. Our results have shown that ML is an effective aid in formulating plant-based meat analogs, laying out the groundwork to expediently optimize product development cycles to reduce costs.

Nonenzymatic Browning of Amorphous Maltose/Whey Protein Isolates Matrix: Effects of Water Sorption and Molecular Mobility

Nonenzymatic browning (NEB) reactions often affect the nutritional quality and safety properties of amorphous food solids. Developing a proper approach to control the NEB reaction has been of particular interest in the food industry. An NEB reaction in an amorphous maltose/Whey protein isolates (WPI) matrix containing L-lysine and D-xylose as reactants were studied at ambient temperatures aw ≤ 0.44 and 45~65 °C. The results indicated that the presence of NEB reactants barely disturbed the water sorption behavior of the matrix. The Guggenheim–Anderson–de Boer (GAB) constants and Qst values of the studied samples were affected by storage conditions as the migration of sorbed water among monolayers occurred. The rate of color changes and 5-hydoxymethylfurfural (5-HMF) accumulation on the matrix were accelerated at high ambient temperatures aw, reflecting the extent of NEB reaction increases. Since the strength concept (S) could give a measure of molecular mobility, the extent of the NEB reaction was governed by the molecular mobility of the matrix as the activation energy (Ea) of 5-HMF production minimized at solids with high S values. We found that the S concept had a considerable potential usage in controlling the NEB reaction on amorphous sugar–protein solids. This data set has practical significance in the comprehensive understanding of manipulating the diffusion-limited chemical reactions on low-moisture food solids.

Machine Learning Model for Quality Parameters Prediction and Control System Design in the Kecombrang Flower (Etlingera elatior) Extraction Process

Kecombrang flowers have bioactive components that can be used as food additives. The development of the kecombrang functional food industry for the production of food additives requires information on production parameters. The extraction process for kecombrang to obtain bioactive components, especially phenols and flavonoids, requires maximum temperature treatment and extraction time. This study aims to determine the standard for the kecombrang flower extraction process, create a machine learning model to estimate the quality parameters of the extraction results (phenol, flavonoid, pH, color, and viscosity), and design a strategy for controlling the extraction machine work to maintain the quality of the extraction, especially of phenols and flavonoids. This research was conducted at extraction temperatures of 60 °C, 65 °C, 70 °C, and 75 °C. During the extraction process, the quality of the material was checked by measuring phenol and flavonoid contents, as well as color, pH, and viscosity. Sampling was carried out at 5 min intervals. The data on the quality parameters during the extraction process were analyzed for trends. A machine learning model, which is an artificial neural network, was developed using a 2–6–1 architecture for each quality parameter. The two inputs of ANN were temperature of extraction and extraction time (duration). The output was the quality parameters of the products (phenols, flavonoids, pH, viscosity, and color), which were evaluated separately. The results show a good correlation between the model and the experimental data, with both the training dataset and the testing dataset. These results were then used to formulate a strategy for controlling the extraction process. A neuro-control system was used as a strategy. This control system was adaptive to changes that occurred during the extraction process so that phenols and flavonoids could be maintained.

An Efficient Automatic Fruit-360 Image Identification and Recognition Using a Novel Modified Cascaded-ANFIS Algorithm

Automated fruit identification is always challenging due to its complex nature. Usually, the fruit types and sub-types are location-dependent; thus, manual fruit categorization is also still a challenging problem. Literature showcases several recent studies incorporating the Convolutional Neural Network-based algorithms (VGG16, Inception V3, MobileNet, and ResNet18) to classify the Fruit-360 dataset. However, none of them are comprehensive and have not been utilized for the total 131 fruit classes. In addition, the computational efficiency was not the best in these models. A novel, robust but comprehensive study is presented here in identifying and predicting the whole Fruit-360 dataset, including 131 fruit classes with 90,483 sample images. An algorithm based on the Cascaded Adaptive Network-based Fuzzy Inference System (Cascaded-ANFIS) was effectively utilized to achieve the research gap. Color Structure, Region Shape, Edge Histogram, Column Layout, Gray-Level Co-Occurrence Matrix, Scale-Invariant Feature Transform, Speeded Up Robust Features, Histogram of Oriented Gradients, and Oriented FAST and rotated BRIEF features are used in this study as the features descriptors in identifying fruit images. The algorithm was validated using two methods: iterations and confusion matrix. The results showcase that the proposed method gives a relative accuracy of 98.36%. The Fruit-360 dataset is unbalanced; therefore, the weighted precision, recall, and FScore were calculated as 0.9843, 0.9841, and 0.9840, respectively. In addition, the developed system was tested and compared against the literature-found state-of-the-art algorithms for the purpose. Comparison studies present the acceptability of the newly developed algorithm handling the whole Fruit-360 dataset and achieving high computational efficiency.

Machine learning-based modeling in food processing applications: State of the art

Food processing is a complex, multifaceted problem that requires substantial human interaction to optimize the various process parameters to minimize energy consumption and ensure better-quality products. The development of a machine learning (ML)-based approach to food processing applications is an exciting and innovative idea for optimizing process parameters and process kinetics to reduce energy consumption, processing time, and ensure better-quality products; however, developing such a novel approach requires significant scientific effort. This paper presents and evaluates ML-based approaches to various food processing operations such as drying, frying, baking, canning, extrusion, encapsulation, and fermentation to predict process kinetics. A step-by-step procedure to develop an ML-based model and its practical implementation is presented. The key challenges of neural network training and testing algorithms and their limitations are discussed to assist readers in selecting algorithms for solving problems specific to food processing. In addition, this paper presents the potential and challenges of applying ML-based techniques to hybrid food processing operations. The potential of physics-informed ML modeling techniques for food processing applications and their strategies is also discussed. It is expected that the potential information of this paper will be valuable in advancing the ML-based technology for food processing applications.

Color-based clustering algorithm as a novel image analytical method for characterizing maltose crystallinity in amorphous food models

Maltose crystallization affects the processibility and stability of sugar-rich foods. This study introduced a color-based clustering algorithm (CCA) to analyze crystallinity from the images of amorphous maltose/protein models. The XRD and DSC were also implemented in maltose crystallization characterization and validated the CCA analysis. The results indicated that CCA could effectively recognize maltose crystals (R = 0.9942), and amorphous maltose mainly crystallized to anhydrate α-maltose and β-maltose monohydrate according to its morphological aspects measured by CCA, XRD, and DSC. However, protein could change the mechanism of maltose crystal formation by disturbing the mutarotation and recrystallization processes of unstable β-maltose. Besides, maltose crystal formation and crystallinity were governed by molecular mobility as the CCA-derived Avrami indexes changed with the Strength parameter. Compared to XRD and DSC, the proposed CCA can provide a rapid and quantitative measure for maltose crystallinity and has great potential applications in the online detection of sugar crystallization.

Application of mucilaginous seeds (Alyssum homolocarpum and Salvia macrosiphon Boiss) and wheat bran in improving technological and nutritional properties of pasta

In recent years, dietary fibers have attracted a lot of attention as they reduce calories and witness the glycemic index. In this study, wheat bran (WB) and mucilaginous seeds flour (Qodume Shirazi seeds [QSS], wild sage seeds [WSS]) as sources of insoluble and soluble dietary fiber were used for pasta enrichment (50% WB, 45% WB-5% seed flour, and 40% WB-10% seed flour). The cooking properties, microstructural, textural, glycemic index, and sensory properties of pasta samples were evaluated. Fiber ingredients increased moisture content, cooking loss, and ash of pasta samples. In contrast, swelling indexing, optimum cooking time, and water absorption decreased. The samples containing high fiber had a darker appearance with a stiffer structure. Microstructure confirmed the presence of a developed protein matrix in the witness sample. But by substitution of the WB, a heterogeneous and dense network with small and large cells formed. The mucilaginous seed flours (WB-QSS and WB-WSS samples) improved the uniformity of pasta microstructure in comparison with WB sample. WB pasta samples reduced all sensory scores, but adding seed flours had a more noticeable influence on increasing the sensory properties. The presence of QSS and WSS resulted in more starchy and elastic texture. By using mucilaginous seeds flour in the production of high-fiber pasta, the glycemic index decreased more noticeably. This investigation indicates the positive impact of mucilaginous seeds, especially WSS, on pasta sensorial properties, in line with a strong influence on technological characteristics and decreasing the glycemic index. PRACTICAL APPLICATION: This study determined a practical approach to produce high-fiber pasta by applying mucilaginous seeds with the improvement of technological and sensory properties.

Comprehensive study on applications of artificial neural network in food process modeling

Artificial neural network (ANN) is a simplified model of the biological nervous system consisting of nerve cells or neurons. The application of ANN to food process engineering is relatively novel. ANN had been employed in diverse applications like food safety and quality analyses, food image analysis, and modeling of various thermal and non-thermal food-processing operations. ANN has the ability to map nonlinear relationships without any prior knowledge and predicts responses even with incomplete information. Every neural network possesses data in the form of connection weights interconnecting lines between the input to hidden layer neurons and weights of hidden to output layer neurons, which has a significant role in predicting the output data. The applications of ANN in different unit operations in food processing were described that includes theoretical developments using intelligent characteristics for adaptability, automatic learning, classification, and prediction. The parallel architecture of ANN resulted in a fast response and low computational time making it suitable for application in real-time systems of different food process operations. The predicted responses obtained by the ANN model exhibited high accuracy due to lower relative deviation and root mean squared error and higher correlation coefficient. This paper presented the various applications of ANN for modeling nonlinear food engineering problems. The application of ANN in the modeling of the processes such as extraction, extrusion, drying, filtration, canning, fermentation, baking, dairy processing, and quality evaluation was reviewed.HIGHLIGHTS1. This paper discusses application of ANN in different emerging trends in food process.2. Application of ANN to develop non-linear multivariate modeling is illustrated.3. ANNs have been shown to be useful tool for prediction of outcomes with high accuracy.4. ANN resulted in fast response making it suitable for application in real time systems.

Prediction of Pasting Properties of Dough from Mixolab Measurements Using Artificial Neuronal Networks

An artificial neuronal network (ANN) system was conducted to predict the Mixolab parameters which described the wheat flour starch-amylase part (torques C3, C4, C5, and the difference between C3-C4and C5-C4, respectively) from physicochemical properties (wet gluten, gluten deformation index, Falling number, moisture content, water absorption) of 10 different refined wheat flourssupplemented bydifferent levels of fungal α-amylase addition. All Mixolab parameters analyzed and the Falling number values were reduced with the increased level of α-amylase addition. The ANN results accurately predicted the Mixolab parameters based on wheat flours physicochemical properties and α-amylase addition. ANN analyses showed that moisture content was the most sensitive parameter in influencing Mixolab maximum torque C3 and the difference between torques C3 and C4, while wet gluten was the most sensitive parameter in influencing minimum torque C4 and the difference between torques C5 and C4, and α-amylase level was the most sensitive parameter in predicting maximum torque C5. It is obvious that the Falling number of all the Mixolab characteristics best predicted the difference between torques C3 and C4.

Prediction of benzo[a]pyrene content of smoked sausage using back-propagation artificial neural network

BACKGROUND

Benzo[a]pyrene (BaP), a potent mutagen and carcinogen, is reported to be present in processed meat products and, in particular, in smoked meat. However, few methods exist for predictive determination of the BaP content of smoked meats such as sausage. In this study, an artificial neural network (ANN) model based on the back-propagation (BP) algorithm was used to predict the BaP content of smoked sausage.

RESULTS

The results showed that the BP network based on the Levenberg-Marquardt algorithm was the best suited for creating a nonlinear map between the input and output parameters. The optimal network structure was 3-7-1 and the learning rate was 0.6. This BP-ANN model allowed for accurate predictions, with the correlation coefficients (R) for the experimentally determined training, validation, test and global data sets being 0.94, 0.96, 0.95 and 0.95 respectively. The validation performance was 0.013, suggesting that the proposed BP-ANN may be used to predictively detect the BaP content of smoked meat products.

CONCLUSION

An effective predictive model was constructed for estimation of the BaP content of smoked sausage using ANN modeling techniques, which shows potential to predict the BaP content in smoked sausage. © 2017 Society of Chemical Industry.

Trends in Chemometrics: Food Authentication, Microbiology, and Effects of Processing

In the last decade, the use of multivariate statistical techniques developed for analytical chemistry has been adopted widely in food science and technology. Usually, chemometrics is applied when there is a large and complex dataset, in terms of sample numbers, types, and responses. The results are used for authentication of geographical origin, farming systems, or even to trace adulteration of high value-added commodities. In this article, we provide an extensive practical and pragmatic overview on the use of the main chemometrics tools in food science studies, focusing on the effects of process variables on chemical composition and on the authentication of foods based on chemical markers. Pattern recognition methods, such as principal component analysis and cluster analysis, have been used to associate the level of bioactive components with in vitro functional properties, although supervised multivariate statistical methods have been used for authentication purposes. Overall, chemometrics is a useful aid when extensive, multiple, and complex real-life problems need to be addressed in a multifactorial and holistic context. Undoubtedly, chemometrics should be used by governmental bodies and industries that need to monitor the quality of foods, raw materials, and processes when high-dimensional data are available. We have focused on practical examples and listed the pros and cons of the most used chemometric tools to help the user choose the most appropriate statistical approach for analysis of complex and multivariate data.

Optimization of Total Monomeric Anthocyanin (TMA) and Total Phenolic Content (TPC) Extractions from Red Cabbage (Brassica oleracea var. capitata f. rubra): Response Surface Methodology versus Artificial Neural Network

The aim of this study was to investigate the influence of solvent type, ultrasonic frequency, extraction time and temperature on the total phenolic content (TPC) and total monomeric anthocyanin (TMA) extraction from red cabbage (Brassica oleracea var. capitata f. rubra) using the response surface methodology (RSM) and artificial neural network. The red cabbage has been used as TPC and TMA sources due to its low cost and highly availability during all the year. The experimental data for the extraction of TPC and TMA were fitted to second-order polynomial models with higher regression coefficients than 0.902. The optimal conditions (in dry matter) for highest TPC extraction (7,049.5 mg gallic acid equivalent/kg) are: methanol as solvent, 3.60 kHz ultrasonic frequency at 67.6 °C for 59.6 min, while for TMA optimal extraction (0.3 mg/g) 2-propanol was used as solvent, 45 kHz ultrasonic frequency at 69.2 °C for 20.80 min. The artificial neural network (ANN) is better than RSM to predict the TPC and TMA extraction from red cabbage.

Comparison of spectral and image morphological analysis for egg early hatching property detection based on hyperspectral imaging

The use of non-destructive methods to detect egg hatching properties could increase efficiency in commercial hatcheries by saving space, reducing costs, and ensuring hatching quality. For this purpose, a hyperspectral imaging system was built to detect embryo development and vitality using spectral and morphological information of hatching eggs. A total of 150 green shell eggs were used, and hyperspectral images were collected for every egg on day 0, 1, 2, 3 and 4 of incubation. After imaging, two analysis methods were developed to extract egg hatching characteristic. Firstly, hyperspectral images of samples were evaluated using Principal Component Analysis (PCA) and only one optimal band with 822 nm was selected for extracting spectral characteristics of hatching egg. Secondly, an image segmentation algorithm was applied to isolate the image morphologic characteristics of hatching egg. To investigate the applicability of spectral and image morphological analysis for detecting egg early hatching properties, Learning Vector Quantization neural network (LVQNN) was employed. The experimental results demonstrated that model using image morphological characteristics could achieve better accuracy and generalization than using spectral characteristic parameters, and the discrimination accuracy for eggs with embryo development were 97% at day 3, 100% at day 4. In addition, the recognition results for eggs with weak embryo development reached 81% at day 3, and 92% at day 4. This study suggested that image morphological analysis was a novel application of hyperspectral imaging technology to detect egg early hatching properties.