Hindawi

Journal of Food Quality

Volume 2022, Article ID 6302331, 6 pages
https://doi.org/10.1155/2022/6302331

Research Article
An IoT and Machine Learning-Based Model to Monitor Perishable
Food towards Improving Food Safety and Quality

Amina Batool ,1 Souvik Ganguli ,2 Hashem Ali Almashaqbeh ,3 Muhammad Shafiq ,4

A. L. Vallikannu ,5 K. Sakthidasan Sankaran ,5 Samrat Ray ,6 and F. Sammy 7
1
School of Automation, Beijing Institute of Technology, Beijing, China
2
Department of Electrical and Instrumentation Engineering, Thapar Institute of Engineering and Technology,
Patiala 147004, India
3
Okan University, Istanbul, Turkey
4
School of Artificial Intelligence, Neijiang Normal University, Neijiang, Sichuan, China
5
Department of ECE, Hindustan Institute of Technology and Science, Chennai, India
6
Sunstone Eduversity, Gurugram, India
7
Department of Information Technology, Dambi Dollo University, Dembi Dolo, Welega, Ethiopia

Correspondence should be addressed to F. Sammy; [email protected]

Received 12 March 2022; Revised 1 May 2022; Accepted 7 May 2022; Published 6 June 2022

Academic Editor: Rijwan Khan

Copyright © 2022 Amina Batool et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Increased quantities of the same sort of item are not nearly as critical to client happiness as a high-quality product. The re-
quirements and expectations of the consumer have an impact on the overall quality of a product or service. The term “quality” may
also be defined as the sum total of all the features that contribute to the production of goods and services that are satisfactory to the
consumer. Certain imported commodities have lately seen an improvement in quality thanks to efforts by importing nations.
Additionally, it safeguards food imported from other nations by confirming that it is safe for human consumption before it is
released. This article describes a technique for monitoring perishable goods that is based on the Internet of Things and machine
learning. Pictures are recorded using high-resolution cameras in this suggested architecture, and then these images are sent to a
cloud server using Internet of Things devices. When uploaded to a cloud server, these photos are segmented using the K-means
clustering method. Then, using the principal component analysis technique, features are extracted from the photos, and the
images are categorized using machine learning models that have been trained. This proposed model makes use of the Internet of
Things, image processing, and machine learning to monitor perishable food.

1. Introduction primary indicator of their immediate sensory quality.

Product’s quality is generally judged by its appearance,
Things of the same type are not as important to the cus- including its color, texture, size, shape, and imperfections
tomer’s satisfaction as high-quality products [1]. The quality [3]. Food manufacturing company administrators take into
of a product is influenced by the needs and expectations of consideration the product’s significance, the social setting,
the customer. Another way to define quality is to think of it and the difficulties experienced by farmers in accomplishing
as the sum total of all the characteristics that go into gen- their agricultural tasks [4].
erating things that the customer is satisfied with [2]. The discovery of faults prior to the sale or export of items
Importing countries have recently improved the quality of is one of the most essential parts of quality assurance [5].
certain imported items. Additionally, it protects food im- When it comes to determining the quality of mangos, hu-
ported from other countries by ensuring that it is safe for man operators have long relied on their eyes. This is not the
consumption. Agricultural products’ external quality is the case anymore. They take a long time, are tedious, and are
2 Journal of Food Quality

inconsistent [6]. In agricultural engineering studies [7, 8], labeled, while the training data for regression must be un-
this method has been applied. labeled (clustering) [14].
Technological advances and the incorporation of these Identifying the traits that make something unique, or
advances into a sustainable farming system have assisted “target attributes,” is the first step in determining how to
agricultural research. Technology has automated every stage classify it. Decision trees and support vector machines, as
of the agricultural process, making it a long-term economic well as Bayesian classifiers, are commonplace in many fields
model. An intermediary who buys cheap from farmers and [15]. They are used in machine learning as well as in neural
then resells it at a profit is removed from the equation. networks and classifiers. One at a time, they will be covered
Computational intelligence approaches have recently been in detail. Algorithms for learning form the basis of each
used to handle agricultural system decision modelling approach.
challenges. Technologies like artificial intelligence and the Decision trees may be useful in situations when classi-
Internet of Things are already making a difference in ag- fication is a problem. A decision tree sorts events into
riculture. Machine learning and the Internet of Things may categories based on the values of the attributes they contain.
be used to keep tabs on perishable commodities, which is an Each branch in the decision tree indicates a possible value for
essential application. Food safety and quality are improved a node in the classification. Feature values are used to rank
as a result [9]. instances in the hierarchy, whereas root nodes are used to
The Internet of Things (IoT) is a hot issue right now. classify them.
Every element of our everyday life is being influenced by the When it comes to input attributes, it may not be possible
Internet of Things. The Internet of Things has numerous to accurately describe them. While input attribute set values
applications and is making people’s lives better (IoT). IoT match those in the training dataset, class variables are
technology enables smart cities, smart homes, smart nondeterministic. This is due to a lack of consideration for
healthcare, smart agriculture, smart farming, smart logistics, noise and confusion in the data processing. There are many
and smart production [10, 11]. ways to tell if you have cardiovascular disease in your daily
Agriculture is essential to the country’s economy since it life.
feeds its inhabitants. This means that all of the country’s Even those who follow a heart-healthy diet and engage in
major corporations are linked and interrelated. A country is a regular physical activity run the risk of getting heart disease
considered rich if its economy and society are mainly de- if they smoke, drink alcohol, or have a genetic predisposi-
pendent on agriculture. Agriculture employs the vast ma- tion. It is not always possible to get reliable information from
jority of people in the vast majority of countries. The upkeep the well-known symptoms of cardiac disease. In these cases,
of plants and animals on large farms may entail the hiring of the Bayesian classifier uses attributes to determine the class
additional workers. The bulk of these huge farms has close label [16].
processing facilities where they generate and process their In order to create animal brains, a biological neural
agricultural goods. network is used. ANN is a connectionist system due to the
Automated machine learning techniques can be used to fact that it is made up of connected nodes and directed links.
analyze the complex interaction between agricultural sys- Communication between nodes is carried out by means of
tems’ inputs and outputs by employing nonlinear and time- connections, each of which has its own weight. Prior to being
dependent analytical methodologies as well as different transferred to the next node, signals must be processed at
unknown components [12, 13]. each node’s end.
This article presents an Internet of Things and machine Nonlinear functions are applied to the output of each
learning-based model to monitor perishable foods. In this neuron in most artificial neural network implementations.
proposed model, first images are captured using high-res- As the process of learning continues, the strength of the
olution cameras, and then these images are forwarded to a signal changes based on the weight and connections of the
cloud server through IoT devices. On the cloud server, these artificial neurons [17].
images are segmented using the K-means clustering algo- There are two ways to build a learning model in ML
rithm. Then, features are extracted by the principal com- classification. These models begin learning as soon as they
ponent analysis algorithm. After that, the images are have access to the training data. A separate model can
classified using trained machine learning predictors. This recognize a test case if the properties of the test case exactly
proposed model makes use of the Internet of Things, image match any of the training instances. Students like these are
processing, and machine learning to monitor perishable labeled “lazy” [18].
food. When it comes to categorization, each sample is con-
sidered an individual data point in the context of all other
2. Literature Survey samples. A clear separation between the test case and the rest
of the data has been formed. A data point’s neighbors in-
The term “machine learning” (ML) refers to an emerging clude every other data point in its immediate proximity.
subject in which algorithms can learn and develop on their The data point is classified in accordance with the
own without the need for explicit coding. There are two main classifications of its immediate neighbors. Data points
approaches to machine learning: unsupervised and super- having a large number of classified neighbors are given the
vised. Classification and regression are the two main types of class label that is most frequently used by other data points.
inference. The training data for classification must be The precise value of K-nearest neighbors must be
Journal of Food Quality 3

determined in order to complete the task. Using a modest

value for k can lead to incorrect classification due to noise in Fruit Images Acquisition by High Resolution
Camera and IoT Devices
the training data. It is possible that data points that are
outside of the test attribute’s immediate area can be found in
the number of closest neighbors that are greater than k. An
error in classification could be the outcome.
Initially, random forest is a machine learning technique Image Segmentation – K Means
that uses many decision trees constructed from random
vector inputs to produce a forest of judgments. Classification
and regression problems can be solved using this method.
The random forest’s forecasts will be more accurate if the
number of trees in the random forest is increased. It is Feature Extraction using Principal Component
critical to realize that a forest and a decision tree are not Analysis
interchangeable.
For the root node to be found, randomly separating
feature nodes is the fundamental difference between deci-
sion trees and random forests. Because of its advantages, Classification of Images by Machine
random forest classification is frequently employed. One of Learning Predictors
its numerous advantages is that it can be utilized for both
classification and regression analysis. It also avoids the COA-SVM, KNN, Naïve Bayes
problem of overfitting if there are many trees available.
Random forest classification can also be used to describe
categorical data and deal with missing data.
Retail, healthcare, and financial businesses can benefit
from random forest classifiers that can be employed in a range
of industries. When it comes to banking, random classifiers Detection of
Perishable Fruits
are frequently employed to distinguish between legitimate
consumers and potential scammers. Random forest is used in
medicine to find the optimum pharmaceutical combinations Figure 1: Internet of things, image processing, and machine
and to diagnose illness based on the medical history of a learning-based model for monitoring perishable food items.
patient. Random forest is the swankier way of tracking a
stock’s performance and calculating its profit or loss. You may
use random forest to predict what products customers will region-based and edge-based. Region-based segmentation
buy based on their preferences in an online store. may be used to classify anatomical or functional features into
The finest tool for categorizing is the support vector groups based on patterns in intensity values around a cluster
machine (SVM) (SVM). As a result, excitement is high of neighboring pixels.
among the employees. In vector space, it is feasible to Based on the unique textures and patterns seen in each
distinguish between instances of different categories using area, this research uses regional segmentation to split the
an SVM model. When they are mapped into the specified ROI. The local mean is utilized as a clustering pattern to
vector space, new samples are given a label based on which assign each unique observation to one of the K-means
side of the gap they fall into [19]. Nonlinear classification can clusters. This program looks for data clusters based on the
be accomplished using an SVM by utilizing the kernel total number of groups specified by the variable k. On-the-fly
technique. data point proximity is determined by calculating squared
Euclidean distances. Data points are allocated to one of the k
3. Methodology categories according to the input characteristics. In order to
group data points, a similarity measure is utilized [20].
This section presents an Internet of Things and machine Using a white or black backdrop, the number K shows
learning-based model to monitor perishable foods. In this that clustering is effective. But if you take a picture of an
proposed model, first, images are captured using high-res- object with a live backdrop and precisely segment the ROI, it
olution cameras, and then these images are forwarded to a fails to work. K shows that clustering is unable to recover the
cloud server through IoT devices. On the cloud server, these ROI area from a picture with a moving backdrop.
images are segmented using the K-means clustering algo- Classification data is classified using KNN’s basic clas-
rithm. Then, features are extracted by the principal com- sifier, which utilizes the tuple that is closest to the unknown
ponent analysis algorithm. After that, the images are [21]. Using likeness measurements such as Euclidean dis-
classified using trained machine learning predictors. The tance, Manhattan distance, Pearson correlation, and others,
block diagram of the model is shown in Figure 1. it predicts new cases. For categorical values, the hammering
Image segmentation is the process of identifying and distance must be determined and standardized using the
grouping together parts of a picture based on shared numerical variables 1 and 0. For the majority of datasets, a K
characteristics. Two kinds of segmentation approaches are value of 3–10 is ideal.
4 Journal of Food Quality

Accuracy (%) Sensitivity (%)

120
100
90 100
80
70 80
60
50 60
40
40
30
20 20
10
0 0
CAO SVM KNN Naïve bayes COA SVM KNN Naïve Bayes
Accuracy (%) Sensitivity (%)
Figure 2: Accuracy of classifiers for perishable food identification. Figure 3: Sensitivity of classifiers for perishable food identification.

COA-SVM (cuckoo optimization algorithm) support Specificity (%)

vector machine is the best option for nonprobabilistic binary 105
linear classification. One or more target classes can be
identified using this method. A single point is used to 100
represent each piece of data. The clear differences among the
various groupings cause them to expand in width. Those new 95
instances are remapped to different target classes based on
where they fall in the gap. Nonlinear classification is possible
90
when the input datasets are unlabeled. The support vector
machine employs an unsupervised learning technique to
85
classify the data because the instances cannot be allocated to
target classes. More instances are added when the clusters
80
based on functions have been formed. A nonlinear support
vector machine-based recommendation system has been
75
demonstrated to be effective. Nonlinear support vector COA SVM KNN Naïve Bayes
machine approaches are the most frequently utilized way for
dealing with unlabeled data [22]. Specificity (%)
Naive Bayes is a subcategory of Bayes’ theorem. It is a Figure 4: Specificity of classifiers for perishable food identification.
monitored, classified system. A simple approximation of
iterative parameters is not required when building a Naive
Bayesian model [23]. The posterior likelihood of P (c|x) can including 158 of them while they are still fresh and 142 of
be calculated using the Bayes theorem, which takes into them after they are rotting. Oranges are the subject of 200
account the prior likelihood of P(c), P(x), and P (x|c). The photographs. There are 120 photographs of oranges in their
posterior probability can be calculated by first generating a prime and 80 images of decaying mangoes in their worst
frequency table for the desired property. A Naive Bayesian state. The K-means clustering technique is used for image
formula is then used to calculate each group’s posterior segmentation. The fruit photos are then classified using COA
likelihood based on the frequency tables. The category with SVM, KNN, and Naive Bayes machine learning algorithms.
the highest posterior likelihood is the result of the These algorithms determine if a piece of fruit is excellent or
calculation. bad. Accuracy, sensitivity, and specificity are shown in
Figures 2–4.
Parameters used in the experimental study are discussed
4. Results and Discussion as follows:
First, images are captured using high-resolution cameras, TP + TN
accuracy � ,
and then these images are forwarded to a cloud server TP + TN + FP + FN
through IoT devices. On the cloud server, these images are
TP
segmented using the K-means clustering algorithm. Then, sensitivity � , (1)
features are extracted by the principal component analysis TP + FN
algorithm, after that the images are classified using trained TN
machine learning predictors. specificity � ,
TN + FP
For research purposes, 500 images of bananas and or-
anges have been gathered. There are pictures of both healthy where TP � true positive; TN � true negative; FP � false
and unhealthy fruits. Bananas are the subject of 300 pictures, positive; and FN � false negative.
Journal of Food Quality 5

5. Conclusion fresh-cut iceberg lettuce through packaging material,” Journal

of Food Engineering, vol. 223, pp. 46–52, 2018.
Monitoring of perishable food items is an important re- [6] P. Sharma, Y. Paul Singh Berwal, and W. Ghai, “Performance
search area as far as smart agriculture is concerned. It takes analysis of deep learning CNN models for disease detection in
the Internet of Things, image processing, and machine plants using image segmentation,” Information Processing in
learning to accurately monitor perishable food items. In- Agriculture, vol. 7, no. 4, pp. 566–574, 2019.
creased quantities of the same type of item are not nearly as [7] M. Ji, L. Zhang, and Q. Wu, “Automatic grape leaf diseases
critical to client happiness as a high-quality product. The identification via united model based on multiple convolu-
consumer’s needs and expectations influence the overall tional neural networks,” Information Processing in Agricul-
quality of a product or service. The term “quality” can al- ture, vol. 7, no. 3, pp. 418–426, 2019.
[8] X. Guo, X. Zhao, Y. Liu, and D. Li, “Underwater sea cucumber
ternatively be defined as the sum of all the characteristics
identification via deep residual networks,” Information Pro-
that contribute to the production of consumer-satisfactory
cessing in Agriculture, vol. 6, no. 3, pp. 307–315, 2019.
goods and services. Certain imported commodities have [9] L. F. Santos Pereira, S. Barbon Jr., N. A. Valous, and
recently improved in quality as a result of efforts by D. F. Barbin, “Predicting the ripening of papaya fruit with
importing countries. Furthermore, it protects food imported digital imaging and random forests,” Computers and Elec-
from other countries by ensuring that it is safe for human tronics in Agriculture, vol. 145, pp. 76–82, 2018.
consumption before it is released. This article covers a [10] A. Raghuvanshi, U. K. Singh, G. S. Sajja et al., “Intrusion
perishable commodity monitoring system based on the detection using machine learning for risk mitigation in IoT-
Internet of Things and machine learning. In this proposed enabled smart irrigation in smart farming,” Journal of Food
architecture, images are captured with high-resolution Quality, vol. 2022, Article ID 3955514, 8 pages, 2022.
cameras and relayed to a cloud server via Internet of Things [11] V. Hemamalini, S. Rajarajeswari, S. Nachiyappan et al.,
devices. When these photographs are uploaded to a cloud “Food quality inspection and grading using efficient image
server, they are segmented using the K-means clustering segmentation and machine learning-based system,” Jour-
algorithm. Then, using the principal component analysis nal of Food Quality, vol. 2022, Article ID 5262294, 6 pages,
technique, features are retrieved from the photographs, and 2022.
the images are classified using trained machine learning [12] A. Raghuvanshi, U. Singh, and C. Joshi, “A review of various
models. The COA-SVM algorithm has better accuracy in the security and privacy innovations for IoT applications in
monitoring and detection of perishable food items. healthcare,” Advanced Healthcare Systems, Wiley Online
Library, Hoboken, NJ, USA, pp. 43–58, 2022.
[13] V. D. P. Jasti, A. S. Zamani, K. Arumugam et al., “Compu-
Data Availability tational technique based on machine learning and image
The data shall be made available on request. processing for medical image Analysis of breast cancer di-
agnosis,” Security and Communication Networks, vol. 2022,
pp. 1–7, Article ID 1918379, 2022.
Conflicts of Interest [14] K. G. Liakos, P. Busato, D. Moshou, S. Pearson, and
D. Bochtis, “Machine learning in agriculture: a review,”
The authors declare that they have no conflicts of interest.
Sensors, vol. 18, no. 8, p. 2674, 2018.
[15] B. Gulzar and A. Gupta, “DAM: a theoretical framework for
Acknowledgments SensorSecurity in IoT applications,” International Journal of
Next-Generation Computing, vol. 12, no. 3, 2021.
This research work is self-funded. [16] K. Phasinam, T. Kassanuk, and M. Shabaz, “Applicability of
internet of things in smart farming,” Journal of Food Quality,
References vol. 2022, Article ID 7692922, 7 pages, 2022.
[17] A. Kaya, A. S. Keceli, C. Catal, H. Y. Yalic, H. Temucin, and
[1] M. Tripathi and D. D. Maktedar, “A role of computer vision in B. Tekinerdogan, “Analysis of transfer learning for deep
fruits and vegetables among various horticulture products of neural network based plant classification models,” Com-
agriculture fields: a Survey,” Information Processing in Agri- puters and Electronics in Agriculture, vol. 158, pp. 20–29,
culture, vol. 7, no. 2, pp. 183–203, 2019.
2019.
[2] T. Laikko-Roto, S. Makela, J. Lunden, J. Heikkilä, and
[18] J. Wang, H. Yue, and Z. Zhou, “An improved traceability
M. Nevas, “Consistency in inspection processes of food
system for food quality assurance and evaluation based on
control officials and efficacy of official controls in restaurants
fuzzy classification and neural network,” Food Control, vol. 79,
in Finland,” Food Control, vol. 57, pp. 341–350, 2015.
[3] H. Calvo, M. A. Moreno-Armendáriz, and S. Godoy- pp. 363–370, 2017.
Calderón, “A practical framework for automatic food prod- [19] P. Ratta, A. Kaur, S. Sharma, M. Shabaz, and G. Dhiman,
ucts classification using computer vision and inductive “Application of Blockchain and Internet of Things in
characterization,” Neurocomputing, vol. 175, pp. 911–923, healthcare and medical sector: applications, challenges, and
2016. future perspectives,” Journal of Food Quality, vol. 2021, Article
[4] J. Zhou, Z. Yan, and K. Li, “Understanding farmer cooper- ID 7608296, 20 pages, 2021.
atives’ self-inspection behaviour to guarantee agri-product [20] M. N. Reza, I. S. Na, S. W. Baek, and K. H. Lee, “Rice yield
safety in China,” Food Control, vol. 59, pp. 320–327, 2016. estimation based on K-means clustering with graph-cut
[5] D. P. Cavallo, M. Cefola, B. Pace, A. F. Logrieco, and segmentation using low-altitude UAV images,” Biosystems
G. Attolico, “Non-destructive automatic quality evaluation of Engineering, vol. 177, pp. 109–121, 2019.
6 Journal of Food Quality

[21] Z. Deng, X. Zhu, D. Cheng, M. Zong, and S. Zhang, “Efficient

k NN classification algorithm for big data,” Neurocomputing,
vol. 195, pp. 143–148, 2016.
[22] E. Suryawati, H. F. Pardede, V. Zilvan et al., “Unsupervised
feature learning-based encoder and adversarial networks,”
Journal of Big Data, vol. 8, no. 1, p. 118, 2021.
[23] D. Mondal, D. K. Kole, and K. Roy, “Gradation of yellow
mosaic virus disease of okra and bitter gourd based on entropy
based binning and Naive Bayes classifier after identification of
leaves,” Computers and Electronics in Agriculture, vol. 142,
pp. 485–493, 2017.