Application of artificial neural network in smart food processing

Application of artificial neural network in smart food processing

In the introduction of the first part, everyone has a relatively comprehensive understanding of the types of artificial neural network in smart food processing. 

Today I bring you the introduction of the second part of the review entitled "Smart Food Processing: A Journey from Artificial Neural Network to Deep Learning": Smart Food Processing Based on Artificial Neural Network (ANN).

1 Introduction

Artificial neural network  have been used in many fields. 

In the past few years, research work based on artificial neural network has seen an astonishing growth in application and development. 

In these food-based applications, artificial neural network  play a vital role in the processing of fruits, vegetables, juices, wine, olive oil, meat, fish, various grains, and soft drinks.

2. Application of artificial neural network in fruit processing

The estimation of fruit quality by artificial neural network is the most important form of artificial neural network application in fruit processing. 

Artificial Neural network play an important role in the processing of bananas, strawberries, grapes, pomegranates, dates, mulberries and other fruits.

2.1 Application of artificial neural network in banana processing

Processing bananas with artificial neural network, like other fruits, has always been an interesting area. developed a method for artificial neural network modeling of antioxidant activity and phenolic compounds of bananas with different drying treatments. 

In their conclusion, the use of a simple artificial neural network model can accurately distinguish the dry state, variety and exact type of banana samples by predicting the content and antioxidant activity of phenolic compounds.

2.2 Application of artificial neural network in strawberry processing

At present, researchers have prioritized the application of artificial neural network methods for strawberry processing. 

Urruty et al. proposed a new method for evaluating strawberry aroma by solid-phase microextraction gas chromatography (SPME/GC) and ANN method. 

The SPME/GC method used in their study is a well-organized investigative tool that can quickly distinguish the chemical environment of various aroma types of strawberries based on a certain range of elements. 

In addition, Menlik et al. also proposed a method to simulate the freezing behavior of strawberries using artificial neural network. 

The authors use the BP method as the proposed  artificial neural network structure, and based on the experimental results, they concluded that their method has good accuracy and is suitable for predicting the freezing of strawberries during the drying process.

2.3 Application of artificial neural network in coffee cherry processing

Fuentes et al. introduced a method for coffee cherries recognition based on Artificial neural network and machine vision technology, which can reduce the coffee fruit recognition rate, the number of instances, and increase the number of final products for coffee producers. 

At the same time, Fuentes et al. also considered the use of Artificial neural network to classify coffee berries into two categories: ripe and unripe, which is a new qualitative advance.

2.4 Application of artificial neural network in pomegranate processing

Artificial neural network have also been used in the processing of pomegranate products. Sargolzaei and Moghaddam have developed a new method for predicting pomegranate oil using artificial neural network. 

Simulations were carried out with different methods such as BPNN, RBFNN, ANFIS, and the prediction results were compared. 

According to the analysis of the experimental results, compared with RBFNN and ANFIS, BPNN is an effective tool in predicting the extraction rate of pomegranate oil.

2.5 Application of artificial neural network in jujube processing

Jujube is a fruit that is rarely used for artificial neural network processing. Fadel proposed a new method for data classification using PNNs. 

In the study, the authors finally observed better classification accuracy by analyzing the influence of classification characteristics, color accuracy and other factors.

2.6 Application of artificial neural network in mulberries

Mulberry is a fruit that has been processed with many machine vision techniques. Many studies have shown that artificial neural network have good effects on mulberry processing. 

Fazaeli et al. developed a new method to predict the properties of black mulberry juice with the help of artificial neural network. 

The researchers took into account factors such as temperature, concentration, and the root mean square error of the prediction, and used backpropagation learning to train their Artificial neural network model. 

The experimental results show that their proposed model has higher performance and can better predict the physicochemical properties of mulberry juice.

3. Application of artificial neural network in vegetable processing

Numerous research works have demonstrated the effectiveness of Artificial  neural network in addressing the complexities of vegetable quality estimation and determination. 

However, this study only covers some papers related to vegetable processing. 

In this subsection, the application of these main artificial neural network in vegetable processing will be introduced.

3.1 Application of artificial neural network in potato processing

The method of processing potatoes using artificial neural network has received great attention, and other methods have also been applied to processing potatoes. 

A method, such as that proposed by FL. Anderson et al., uses statistical and artificial neural network classifiers for trace metal analysis to determine the geographic origin of potatoes. 

When considering the effect of seasonal factors on its results, by using the bagging method in the artificial  neural network, a very accurate classification result can be obtained and divided into two potatoes of different seasons. 

Finally, the researchers found that seasonal changes did not affect the classification results of the classification model, further demonstrating the reliability of the method.

3.2 Application of artificial neural network in carrot processing

Erenturk proposed a new technique for carrot drying using an artificial  neural network approach. And the authors compare their proposed method with a genetic algorithm. 

Its main purpose is to solve the drying characteristics of carrots under different aeration conditions. 

The artificial  neural network technique introduced in their paper excels in predictive studies of drying dynamics. 

Their study demonstrates that the capabilities of artificial neural network can be successfully used for drying process control and online condition assessment in the food industry.

3.3 Application of artificial neural network in tomato processing

Vegetables like tomatoes have an immersive artificial neural network processing approach to them. Hahn et al. introduced the spectral detection and artificial neural network identification technology of Rhizopus red tomato spores. 

According to the analysis of the experimental results, they obtained a standard hit rate of 82.62% in the discrimination study using four wavelength relationships, and finally their study confirmed the possibility of using optical reflectance to identify Rhizopus conidia on the ripe fruit of red tomato .

3.4 Application of artificial neural network and image processing in food processing

Artificial neural network for food processing using image processing technology have attracted the attention of researchers and practitioners. 

This article will introduce some applications of artificial neural network in food image processing. 

Ram et al. developed a new method for predicting olive oil content based on image processing. Among them, the researchers found that, in their approach, the articial neural network gave reliable results of an average linear correlation of olives. 

When they used the image features of two-sided olives, its performance was even stronger.

To classify the shape of cooked shrimp, Poonnoy et al. developed an ANN image analysis technique. 

Their experimental results show that their proposed method has a high accuracy of 99.80%, which means that the error rate is very low. 

They considered the RID value to be the characteristic shape of boiled shrimp and showed that hybridizing RID with an artificial neural network improved the performance of their ANN model with high accuracy.

At the same time, various other artificial neural network and image processing are also used to process other foods such as meat, grains, dairy products, and bakery products.

3.5 Application of artificial neural network in olive oil processing

Over the past two decades, many articles have been published on olive oil processing, as shown in Table 1. Peres et al. proposed a new method to classify the fruits of different olive varieties using artificial neural network and linear discriminant analysis. 

In their proposed method, compared to earlier research work, artificial neural network, as an effective and powerful tool, are also used as a tool to prevent adulteration of olive oil and olives of unsuitable varieties. 

Not only that, but a further study by Carfagni et al. in 2008 proposed a new technique for rapid, involuntary prediction of olive oil, and Garcia-Gonzalez et al. proposed a new method utilizing MLP. 

In addition, Gonzalez-Fernandez et al. provide a critical review of the use of ANNs in olive oil production and characterization.

Table 1 Application of artificial neural network in olive oil processingSmart Food Processing: A Journey from Artificial Neural Network to Deep Learning 

3.6 Application of artificial neural network in fish processing

Artificial neural network in fish processing are widely used. 

Lae et al. developed a new artificial neural network method for predicting fish production in African lakes. 

They compared the method with traditional statistical methods and mentioned some limitations of artificial neural network in their paper. 

Furthermore, they pointed out that using BP procedures with artificial neural network has advantages in predicting fish yield. 

In 2003, Maravelias et al. also proposed a new method for predicting Mediterranean fish species using ANN, and compared artificial neural network with discriminant function analysis (DFA). 

After considering the accuracy, sensitivity and specificity, it is finally concluded that ANN is better than DFA, their research became a classic example of the accuracy and ability of artificial neural network  to predict fish species.