ORIGINAL RESEARCH

published: 12 January 2022

doi: 10.3389/fpubh.2021.816226

An Artificial Intelligence Approach

Toward Food Spoilage Detection and
Analysis
Ekta Sonwani 1 , Urvashi Bansal 1*, Roobaea Alroobaea 2 , Abdullah M. Baqasah 3 and
Mustapha Hedabou 4
1
Department of Computer Science and Engineering, Dr. B. R. Ambedkar National Institute of Technology, Jalandhar, India,
2
Department Computer Science, College of Computers and Information Technology, Taif University, Taif, Saudi Arabia,
3
Department of Information Technology, College of Computers and Information Technology, Taif University, Taif, Saudi Arabia,
4
School of Computer Science, Mohammed VI Polytechnic University, Ben Guerir, Morocco

Aiming to increase the shelf life of food, researchers are moving toward new
methodologies to maintain the quality of food as food grains are susceptible to spoilage
due to precipitation, humidity, temperature, and a variety of other influences. As a result,
efficient food spoilage tracking schemes are required to sustain food quality levels. We
have designed a prototype to track food quality and to manage storage systems at home.
Initially, we have employed a Convolutional Neural Network (CNN) model to detect the
Edited by: type of fruit and veggies. Then the proposed system monitors the gas emission level,
Celestine Iwendi,
University of Bolton, United Kingdom humidity level, and temperature of fruits and veggies by using sensors and actuators to
Reviewed by: check the food spoilage level. This would additionally control the environment and avoid
Ebuka Ibeke, food spoilage wherever possible. Additionally, the food spoilage level is informed to the
Robert Gordon University,
customer by an alert message sent to their registered mobile numbers based on the
United Kingdom
Michael Edeh, freshness and condition of the food. The model employed proved to have an accuracy
Ebonyi State University, Nigeria rate of 95%. Finally, the experiment is successful in increasing the shelf life of some
Mohammad Shah,
University of Bolton, United Kingdom categories of food by 2 days.
*Correspondence: Keywords: machine learning for health, smart system, food spoilage detection, food spoilage prevention,
Urvashi Bansal sensors, IoMT
urvashi@[Link]

Specialty section: 1. INTRODUCTION

This article was submitted to
Digital Public Health, Food wastage has been a topic of concern in recent years, and studies are being conducted to
a section of the journal identify innovative ways to reduce it. It has been described as a major concern in the long-term
Frontiers in Public Health
sustainability of food production, demand, and food supply chains. Since meals are the essential
Received: 16 November 2021 source of diet for all living beings, the quality and security of meals have always been in high
Accepted: 14 December 2021 demand. The Internet of Things (IoT) links anything, everywhere, and at any time (1–4) and by
Published: 12 January 2022
incorporating the IoT into the Food Supply Chain (FSC) management, it is possible to improve
Citation: food shelf life by measuring and monitoring the state of the food and exchanging data to and
Sonwani E, Bansal U, Alroobaea R,
from customers. Currently, the whole use of IoT technologies in FSC is only in its early stages,
Baqasah AM and Hedabou M (2022)
An Artificial Intelligence Approach
with a long way to go (5, 6). To avoid food wastage, food sanitation and safety are of paramount
Toward Food Spoilage Detection and importance. The consistency of meals should be regulated. As a result, wastage can be reduced
Analysis. by installing quality control systems in grocery stores. It will further help in controlling diseases
Front. Public Health 9:816226. (7). These types of quality control systems keep an eye on the environmental conditions that may
doi: 10.3389/fpubh.2021.816226 maintain the food quality. Previously, atmospheric effects were monitored by procedures such

Frontiers in Public Health | [Link] 1 January 2022 | Volume 9 | Article 816226

Sonwani et al. Food Spoilage Detection and Analysis

veggies. But everyone cannot identify which fruit or vegetable

is hybrid or not; that is why everybody needs to know the
knowledge of their routine work.
Food poisoning can be caused by a variety of factors,
including humidity and temperature fluctuations. As a result, it
is important to provide a measurement device that can measure
humidity and temperature differences during food preparation
and transportation (10). Currently, nearly everyone is influenced
by the foods they consume daily, not only because of junk food,
but also because of canned vegetables and other food items
eaten in daily life, which lack consistency because their oxygen,
temperature, and moisture content differ from time to time. In
FIGURE 1 | Food spoilage process.
the smart house, devices are installed which have the capacity to
diagnose wastage of food and then alert caregivers. The study
and execution of repetitive measurements, which are aimed
as refrigeration, vacuum storage, and so on. Food pollution at detecting improvements, do not guarantee the nutritional
can occur during the manufacturing process, although it is content of the food (11). The details gathered by tracking
most often caused by inefficient food handling activities due to and monitoring should be reviewed and properly submitted
unsuitable environmental conditions during food transportation to the administrator for purposes of policy research, pattern
and storage. forecasting, program assessment, and planning.
The expanded use of smart phones, connected devices, and
1.1. Food Spoilage Process controllers in food production and other domains has wreaked
Food spoilage occurs when a food product becomes unfit for use havoc across the globe. The advantages of IoT technology are
by the customer. The factors of such a process are a result of a rapidly being tapped by the food industry and other sectors (12–
variety of external factors, including the type of food item, how 14). It accesses via broadband (modem) and begins scanning
it is packaged and processed. Every year, one-third of the world’s input from the wired up to sensors, the heat, and humidity sensor
melded are lost which is provided for human consumption due to until properly mounted and turned on. The temperature and
meal wastage (8, 9). In Figure 1, it could be visualized that once moisture controller sensor is a wireless sensor with an active
the germs start attacking fresh apples, gas emission takes place moisture controller sensor. We have employed this idea in the
and after few hours, the apple is spoiled completely. proposed scheme.
Bacteria, virus, protozoa, and fungi are factors of food Again, the primary function of the proposed monitoring and
spoilage. These factors can create harmful results for consumers, control systems is to keep track of a certain operation and ensure
but we can apply prevention techniques to them to save the that it continues as desired. The surveillance process can be
life and quality of food. Usually, bacteria can not cause food accomplished using a variety of different types of sensors. The
poisoning and most of the microorganisms that cause food borne data collected from the Arduino-based sensors will be compared
illness are odorless and tasteless, otherwise owing to mycotoxins with the target values (15). If the reading value of the sensor is
and microbial wastes. Therefore, eating spoiled food is never found to be different from the target values, the control circuit
recommended. Two pathogenic bacteria Clostridium perfringens will intervene to affect the allocated operation to keep it at the
and Bacillus cereus can spoil the food. desired level of quality. As a result, the Smart food spoilage
monitoring system focuses primarily on healthy food storage
1.2. Key Component of Food by surveilling and regulating a variety of parameters that affect
There is a very important key component i.e., seed. If we want food products. This control system employs storage units that
healthy fruits and veggies, that seed must be strong and healthy. are equipped with a variety of sensors that read the parameters
The fruits and veggies completely depend on the seeds. There are influencing food quality (16).
two types of seeds:
• Hybrid seeds: Cross-pollinated plants produce hybrid seeds in
1.3. Reasons for Food Spoilage
The recycling bins and the waste made are evidence of the
cultivation and gardens. Hybrids are selected for their ability
food spoilage circumstances. The big reason for food spoilage
to enhance the characteristics of the resulting plants, such as
is that a larger portion of food waste occurs in office eateries,
yield, uniformity, color, and disease resistance.
small and large roadside eateries, group get-together events,
• Non-hybrid seeds: These seeds are called heirloom seeds
and wedding receptions. These food surpluses are not only a
or free pollinated seeds. Non-hybrid seeds originate from
widespread indicator of toxicity to the earth’s atmosphere, but
naturally pollinated trees. Any of these cultivars have been
they also pose a slew of economic problems. According to current
around for millennia.
calculations, half of all of these foods are lost globally; the global
The hybrid seeds give hybrid fruits and vegetables and non- amount of meal wastage is estimated as 1.3 billion tons, and it
hybrid seeds give non-hybrid fruits and veggies. The lifetime of is projected to rise even more in the upcoming year, which is a
hybrid fruits and veggies is less than the non-hybrid fruits and concerning problem (17).

Frontiers in Public Health | [Link] 2 January 2022 | Volume 9 | Article 816226

Sonwani et al. Food Spoilage Detection and Analysis

FIGURE 2 | Factors of food spoilage.

Food spoilage can be caused by a variety of human, chemical, • Designing an alert generation system to alert the user about
and biological causes, including plants, enzymes found in the quality of food and the time spans of its deterioration.
plant food tissues, insects, parasites, and microbes (18). The • Development of a system to increase shelf life of food by
most common and serious cause of food spoilage is microbial maintaining the environment according to food.
degradation. Food will invariably be polluted by such forms
and amounts of microorganisms when microbes are commonly 2. LITERATURE REVIEW
spread in nature.
Figure 2 shows the factors that can destroy fruits and veggies. Continuous sensing in a smart home is a unique sensor modality
In terms of heat, humidity, moisture, temperature, and oxygen that has the ability to distinguish diverse everyday activities and
content they have a threshold value. If the threshold value crosses indicate potentially dangerous situations for the person who
then food spoiling starts. If we maintain a threshold value then lives there. One such situation is food rotting, which is the
we can vary the life of fruits and vegetables. subject of this study (19). The authors have evaluated the smell
signatures of two typical meals (milk and yogurt) that were
kept at 25◦ C for a week using a metal-oxide sensor (MOS)
1.4. Motivation based electronic nose. The highest absolute sensor responses were
The motivation behind Monitoring and analysis of food spoilage used to create feature vectors, and their components followed a
using Machine Learning is to keep track and manage food smooth trajectory as the data aged. The two chemicals followed
products in order to avoid spoilage caused by climatic and different trajectories during spoiling, according to principal
atmospheric changes. Food wastage concerns are focused on component analysis (PCA).
healthy eating habits and also concerns with the quality of food Rajesh Megalingam et al. introduce a unique method for
as there have been instances where poisonous chemicals have detecting food deterioration by combining picture classification
been discovered in fruits and veggies. Monitoring and analysis with machine learning techniques and artificial intelligence
of food spoilage using Machine Learning saves time and provides (20). They have used AI, deep CNN networks, computer
accurate and consistent results. vision, and ML techniques such as the k clusters method
for color classifications in pictures and its HSV values for
spoiling detection to identify food rotting. The anaconda
1.5. Our Contribution
prompt is used to complete this project on the jupyter
While reviewing the articles on food wastage, we have found that
notebook platform. Also, Iwendi et al. employed an Artificial
there are a few contributions to increasing the shelf life of food.
Intelligence approach for detecting and analyzing security
To this end, we have found the research questions for this field.
levels on IoT using a network classifier. The study shows
The major and novel contributions of this study are:
an impressive accuracy rate for the proposed application. In
• Designing a smart system to keep track of the quality of food this view, we can say that AI is useful in almost every
using gas, humidity, and temperature sensing mechanism. field nowadays.

Frontiers in Public Health | [Link] 3 January 2022 | Volume 9 | Article 816226

Sonwani et al. Food Spoilage Detection and Analysis

Green et al. presented an electronic nose (e-nose) that consists The electronic nose system has a 3.85% mistake rate, according
of four gas sensors made of functionalized single-walled carbon to the confusion matrix.
nanotubes (f-SWNTs) and polymer nanocomposites, with the Benjamin et al. described a flexible UHF RFID sensor for
goal of becoming a simple monitoring device for microbiological detecting food quality (26). This sensor is based on inter-digital
spoilage and pollutants in canned food (21). The gas sensing capacity found in an RFID antenna, onto which a coating of
signals were utilized as early indications of deterioration to assist vegetal biopolymer has been placed. Electromagnetic coupling
in averting negative health impacts. In a static environment, between the capacity and the biopolymer is, therefore, employed
the gas sensor array was tested to see how it responded to to vary the adaption coefficient between the chip and the antenna
different volatile organic compounds (VOCs). Because they have of the RFID tag based on food deterioration. Experimental
an excellent sensing response to ammonia, which is one of the measurements of an RFID-sensor exposed to a genuine food gas
gases generated by microorganisms. These sensors are suitable environment in the process of deterioration.
for detecting microbial deterioration of canned food. The sample, Food losses after harvest can range from 10 to 30%, owing
which was canned tuna in mineral water, was opened and kept at to a variety of fungal and other microbes that degrade the
room temperature (25◦ C). The odor linked with the deterioration food during transportation and storage (27). In addition to
of canned tuna was then monitored for 10 days using e-nose. PCA these direct losses due to spoiling, several bacteria generate
was also used to visualize the discrimination of microbial canned natural poisons that make foods unsuitable for human ingestion.
tuna spoilage status and analyze the smellprint of a specific level Other fruits and vegetables require waxes and water-loss
of ammonia contamination. PCA was found to be able to track barriers on the surface to retain quality and ideal sensory
changes in canned tuna deterioration, demonstrating its potential characteristics between farm and market. This study describes
to be used for quality assurance of canned food in the daily life of a study and development of an effective electrostatic spray
a smart home resident. application technology as well as a processing-line prototype
When food is reversed in an inadvertent situation, such as an designed particularly for food postharvest protection. The study
insufficient temperature, foodstuff rots as a result of the rapid provides experimental findings of comprehensive assessments
breeding of food spoilage bacteria in warm and wet conditions of electrostatically applied protective sprays onto bananas for
where bacteria can easily be generated (22). The consequences international transport, where both microbiological and mass-
can be even worse if humans eat those degenerative foodstuffs, transfer data show usually 2.1–3.4-fold deposition benefits for
as they may develop bromatoxism. The goal of this study is to food protection.
design and simulate a wearable RFID patch for food spoilage Microbial deterioration in milk is a major source of concern
monitoring with smart packaging that can be recognized and for those who want to eat healthily (28). The goal of this research
read temperature information by a device that supports near NFC is to use an electronic nose system (e-nose) with a variety of gas
technology via an attached circular antenna. sensors as an initial tool for monitoring food deterioration in
Everyone has a distinct opinion on whether or not a meal order to preserve food safety and human health. Nanoparticles
is rotten, which can lead to incorrect conclusions about the gas sensors made of various polymers and f-SWCNTs are used in
item’s state (23). Foodborne sickness might result from a this device. When these gas sensors are exposed to the headspace
misinterpretation of the food condition. Due to this forecast of the milk, they respond to the VOCs present in the milk.
ambiguity, there is still a desire for an electronic nose system Pasteurized milk was stored at 4 and 25◦ C for 9 days to monitor
to categorize if a meal is rotten. The goal of this study is to the development of spoiling. The odor pattern was examined
use an electronic nose to identify food deterioration in tomato- using PCA to assess the capabilities of an electronic nose for
based Filipino cuisines. This study seeks to create a device freshness and milk spoiling detection. Monitoring the change in
with an array of sensors that can detect gases released by detecting responses revealed each sensor’s capacity to identify the
rotten tomato-based Filipino cuisines and to use an Artificial strength of the odor level. The findings of the discrimination
Neural Network as an algorithm to classify the sensor’s data demonstrate that the odor levels of samples maintained at 4◦ C
readings. The e-nose system’s hardware includes a Gizduino do not change from the first to the ninth day, but the odor
1281, a Raspberry Pi 3 Model B, a 20x4 LCD screen, seven levels of samples held at ambient temperature increase with
MQ gas sensors, and one temperature/humidity sensor. The storage duration.
Artificial Neural Network data (24, 25) is trained using Stochastic Food planning and quality assessments are becoming
Gradient Descent and the Back Propagation technique. The increasingly essential aspects of modern society (29).
sensor chamber is positioned underneath the tomato-based We may effectively analyze crop quality utilizing various
Filipino food for simple detection of the gas sensors generated learning approaches thanks to the scientific vision and
by the cuisine. The research might be beneficial in determining agricultural advancements, which certainly replace inconsistent,
whether or not food has gone bad. This electronic nose device unpredictable, and time-consuming human labor. The use of
may detect the degree of rotting in a certain tomato-based fuzzy logic to crop quality has effectively demonstrated its worth
Filipino dish. The researchers assigned rotting levels ranging and established itself as a viable alternative to automation,
from 0 to 12. Starting at 7:00 a.m. on Day 1, these levels correlate indicating that it has a bright future ahead of it. This article
to the period when the meal is monitored every 4 h. As the trial examines some of the most important research findings on
progressed, the food deteriorated often between levels 5 and 6. this subject.

Frontiers in Public Health | [Link] 4 January 2022 | Volume 9 | Article 816226

Sonwani et al. Food Spoilage Detection and Analysis

Food safety is currently a key problem for the food business

(30). Food businesses must develop quality monitoring systems
in order to discover food quality issues early in the production
process. Data manipulation and centralized storage, on the
other hand, have become obstacles to dependability in a typical
quality monitoring system. Furthermore, traditional quality
monitoring techniques are typically ineffective due to a lack of
effective automation. Fortunately, block-chain is a promising,
tamper-proof, and decentralized technology. Furthermore, smart
contracts, which are self-executing and self-verifying programs
on the blockchain network, may execute transactions between
mutually untrustworthy parties. This article offers an intelligent
quality monitoring system for fruit juice manufacturing that
combines smart contracts and quality rating algorithms. This
system offers a high level of automation as well as a high level
of dependability. Response surface models are created based
FIGURE 3 | Food preservation techniques.
on preproduction data in this system, and the best production
condition for each step is determined. Smart contracts are used
to capture production data on a blockchain during the actual
production process. allows you to enjoy a wide range of foods all year long. The four
Food is extremely essential in our day-to-day lives. The quality techniques for food preservation are shown in Figure 3 (32).
of food is deteriorating day by day as globalization progresses.
To keep the food fresh, various food processing is done most of 3.2. Alternatives of Artificial Preservatives
the time. Various preservatives or substances are added to the Various natural preservatives derived from plants, insects, fungi,
meal to give it a fresh or appealing appearance. The majority and alloy have antioxidant, antimicrobial, and anti-enzymatic
of food is now preserved using chemicals, resulting in food effects in studies. Neem, basil, rosemary, and clove take out are
contamination. As a result of the pollution, numerous illnesses encouraging alternatives for their synthetic equivalent (33).
develop, prompting consumers to seek out healthier foods. Studies show that manufacturing companies are using
Organic food is desired by the public in order to maintain a chemical antioxidants which is the cause of hyperactivity in
healthy lifestyle. So, in order to prevent difficulties linked with previously non-hyperactive people. Natural compounds derived
food that cannot be interpreted by humans, we need a gadget that from plants, organisms, or alloy may be useful substitutes. They
can assist in determining the quality of food. There is a need for may even be found as preservatives in meats, cosmetics, and
such a device that instructs us on how to eat sanitary meals. As a pharmaceuticals, where they could be used as flavoring, binding,
result, in order to meet this customer need, we created a gadget disintegrating, gelling, thickening, or suspension agents, even
that determines if the food quality is good or terrible. This article as automobiles.
examines the application of several sensors in the food business. Researchers also used extracts from natural sources such as
Sensors such as pH sensors, gas sensors, and temperature sensors green tea, grape seed, and nisin, a bacteriocin known as a
aid in determining the state of food. This system has a strong nutritious food preservative, to make chicken and turkey hot
presence in restaurants, homes, and small businesses (31). dogs in a series of trials (34). Table 1 represents the information
regarding various food preservatives used for different kinds of
fruits and vegetables. Table 2 represents the diseases caused using
3. EXISTING METHODS AND SOLUTIONS preservatives.
The goal of the smart meal tracking device is to track and 3.3. Importance of Food Preservation
regulate food products to avoid harm caused by atmospheric or • Bacterial development and other forms of spoilage are
climate variability. Food waste may also be caused by insufficient inhibited by preservation processes, ensuring that the food is
storage of foods. Through tracking and regulating different healthy enough to consume.
criteria concerning foodstuffs, the smart food monitoring • Pickling competes with freezing, canning, and drying as a way
system focuses on healthy food storage. This device employs to save produce from spoiling. Antioxidants, amino acids, and
storage units that have been implanted with a variety of beneficial bacteria are commonly found in fermented foods.
electronic sensors that can interpret the parameters that • Food preservation delays the breakdown of rancid-causing
affect foodstuffs. fats while also inhibiting the development of microorganisms
(such as yeasts) and other microorganisms.
3.1. Techniques for Food Preservation • Feed storage expands the stock of food.
Food preservation’s main goal is to save food from spoiling so it • Feed storage helps to reduce food waste. Excess foods that
can be eaten. Gardeners also grow too much food at once, much would have been discarded otherwise are processed and stored,
more than can be consumed until it spoils. Food preservation also adding to available stocks and reducing food waste (35).

Frontiers in Public Health | [Link] 5 January 2022 | Volume 9 | Article 816226

Sonwani et al. Food Spoilage Detection and Analysis

TABLE 1 | Different types of food containing various kinds of preservatives. 4. ARCHITECTURE

Preservatives Food The proposed solution includes new device architecture with
few configuration modifications as per the requirement. The
Sorbic acid Syrups, sweets, dairy products, fruit
products, fermented products, beverages new device architecture includes a humidifier to maintain the
Tert butyl hydroquinone (TBHQ) Fats, oils, snack foods humidity of the device and a cooling module that maintains
Tocopherols (vitamin E) Oils the temperature for the food items. The prototype of the device
Ascorbic acid (vitamin C) Fruit and acidic products
is represented in Figure 4. Various electronic instruments have
Butylated hydroxyanisole (BHA) and Fats and oils, bakery products, cereals
been used for monitoring. Furthermore, the registered values are
Butylated hydroxy -toluene (BHT) used for monitoring purposes (36). The information gathered by
Sodium sorbate Mayonnaise, processed meats, dairy various sensors can be correlated to the target values. The speed
products, fermented products control kicks in and interprets the provided procedure to keep
Sodium and calcium propionate and Breads and other baked goods it running in the correct direction if the sensor readings do not
Potassium propionate and propionic match the key parameters. This theory can be used to create a
acid method that can protect raw foods. This architecture consists of
Benzoic acid and sodium benzoate Fruit products, margarine, and acidic the following components:
foods
Calcium lactate Olives, frozen desserts, jams, jellies, and • Controller: Arduino is used as a microcontroller.
dairy products • Gas detection sensor: A gas detector is a device that measures
Calcium sorbate Mayonnaise, dairy products, syrups, and the number of gases in a given environment, and is often used
margarine as part of a protection system. Operators in the field where the
Ethylene diamine tetra acetic acid Dressings, canned veggies, and leak is happening will be alerted by a gas detector (37, 38).
(EDTA) margarine
• Humidifier: A humidifier is a system that raises humidity
Methylparaben Relishes, dressings, and beverages
(moisture) in a single room or a whole house. It is essentially
Propylparaben Cake, pastries, beverages, and relishes
an electrical appliance. Point-of-use humidifiers are widely
Sodium nitrate and nitrite Cured meats, fish, and poultry
used in the home to humidify a single room, and humidifiers
are used in medical ventilators for greater patient convenience.
• Heat sensor: The key function of a heat sensor is to detect the
heat that is present inside it. When the temperature around the
• Dietary shortages can be reduced by storing food. Preserved heat sensor rises above its fixed point, it detects the heat and
foods help to bring diversity to the diet. In certain Middle alerts us, allowing us to protect the devices from injury (39).
East nations, there is a shortage of plantations due to arid • Humidity sensor: A humidity sensor is a sensor that senses
environmental conditions. Importing new and dried fruits and humidity and converts the information into an electrical
veggies makes up for this shortcoming. current. Humidity sensors are available in a range of shapes
and sizes, as well as with different features (40). The Humidity
sensor first senses any humidity or moisture in the fruits or
vegetables and it sends some value to the microcontroller if
3.4. Reasons for Food Preservation
any moisture content is detected then an alert message sends
• The product could not be stored for a longer time as it could
to our mobile phone if the moisture content is not detected
be spoiled.
then again taking values from the microcontroller.
• Using a new product that is in season and produced locally,
• Cooling module (TEC1-12715-Thermoelectric Cooler 15A
remaking preserving jars, cutting food waste, and reducing
Peltier Module): The heater, condenser, and fan device are
food miles are all eco-friendly choices.
all included in this element. A heater and a condenser are
• Buy in bulk and seasonally to save costs.
included in this module.
• Artificial preservatives, BPA, and other artificial additives may
• Light sensor: Light sensors are electronic devices that monitor
not be used in home-prepared food.
the intensity of natural or artificial light. Light energy is
• Convenience - in minutes, it is ready to eat.
converted into an electrical signal by these devices. Light
• Both the canning process and sharing with families, colleagues,
sensors are used in a variety of manufacturing application
and neighbors need community participation. Teach the next
areas.
generation how to do things.
• Have plans for unexpected events and/or injury. In Figure 5, the Gas detection sensor senses any gas or air coming
• Personal fulfillment, we enjoy coming up with different flavor from the fruits or vegetables, and it sends some value to the
blends that you would not find in your grocery supermarket, microcontroller if any gas content is detected then it alerts the
as well as demonstrating how to use packaged foods in recipes. user. If gas content is not detected then again values are taken
• Learn new skills, such as how to use all of the fruit and how from the microcontroller.
to cut down on food waste. Consider the following scenario: In Figure 6, the Humidity sensor first senses any humidity or
Apples - cut into slices (or use as pie filling) and squeeze the moisture in the fruits or vegetables, and it sends some value to the
juice from the peels and cores. microcontroller. If any moisture content is detected then it alerts

Frontiers in Public Health | [Link] 6 January 2022 | Volume 9 | Article 816226

Sonwani et al. Food Spoilage Detection and Analysis

TABLE 2 | Dangerous food preservatives cause various diseases.

Preservatives Cancer possibility Asthma possibility Hypersensitivity

(Yes/No) (Yes/No) possibility (Yes/No)

Calcium/Potassium/Sodium propionate and propionic acid No Yes Yes

Sodium and potassium nitrate Yes No Yes
Sodium nitrite Yes Yes Yes
Butylated hydroxyanisole (BHA) Yes Yes Yes
Butylated hydroxytoluene (BHT) Yes Yes Yes
Tert butyl hydroquinonesynthesiz-ed (TBHQ) No Yes Yes
Sodium benzoate Yes Yes Yes
Potassium and calcium sorbate and Sorbic acid No Yes Yes
Benzoic acid No Yes Yes
Propylparaben No Yes No
Sulfur dioxide No Yes Yes
Potassium bisulfite No Yes Yes
Hexamethylen-etetramine Yes No No
Sodium metabisulphite No Yes No

FIGURE 4 | The architecture of Monitoring and analysis of food spoilage using Machine Learning.

the user and if the moisture content is not detected then again is below the threshold value then the humidifier increases the
values are read by the microcontroller. As shown in Figure 7, the humidity up to the threshold value. The temperature sensor
heat sensor and cooling module senses any heat in the fruits or monitors the temperature for the predefined threshold value
vegetables and it sends some value to the microcontroller. The which is controlled by Arduino. Arduino is directly connected
complete process is given in Algorithm 1. to the Raspberry pi, which acts as a mini computer having own
processing unit and memory. Whenever temperature reaches
above the value of the threshold, the cooling module gets
5. WORKING PRINCIPLE
turned on. The gas sensor detects early spoilage via detecting
We have different types of sensors that play a major role in a little amount of gas emission of the food items. It sends
the detection and monitoring of food spoilage. It includes the the value to the Arduino, after receiving the values, Arduino
camera sensor, humidity sensor, gas sensor, and heat sensor. pushes the values to the cloud. The web application notifies the
The camera sensor captures the image of fruit or vegetable. The user via a buzzer or voice-activated commands or through the
humidity sensor senses the humidity of the environment. If it display messages.

Frontiers in Public Health | [Link] 7 January 2022 | Volume 9 | Article 816226

Sonwani et al. Food Spoilage Detection and Analysis

6. SMART FOOD LIFE PREDICTION Algorithm 1: Process (object).

1: Turn on the device
The food life cycle has different stages which include the 2: Capture the image of fruit or vegetable
following stages: Fresh Food; Food spoilage just started; Spoilage 3: Turn on the Cooling module
Food. For prediction, we have trained 50 types of fruits and 4: Turn on Humidifier
vegetables over different three classes. We proposed a CNN for 5: Store the optimal values of parameter according to captured
object detection and prediction model. This is trained over three object
different classes. 6: Read the values of sensor for monitoring process of fruits or
Object detection is a difficult task in the realm of computer vegetables
vision. Machine learning models when trained on a large 7: if Gas content is detected then
enough number of photos can be used to detect and distinguish 8: go to step 20
objects. We will identify the fruit or vegetable by using the 9: else
Convolutional Neural Network (CNN) to calculate the threshold 10: go to step 21
values for different parameters e.g., gas, temperature, and heat 11: end if
as these values are different for a different type of food. The 12: if Moisture content is detected AND moisture > maximum
networks will be trained in a supervised way, with photos of the optimal humidity of object AND Humidifier is off then
13: Turn on the humidifier
14: else if heat content is detected AND temperature >
maximum optimal storage temperature of object AND
cooling module is off then
15: Turn on the cooling module
16: else if life of object > average shelf life of object then
17: go to step 20
18: else
19: go to step 21
20: end if
21: Send alert to the user
22: Capture an image of fruit or vegetable
23: go to step 7

fruits serving as input and tags serving as a result. The CNN

model will be able to correctly anticipate the fruit label after
successful training.
We have utilized the deep learning approach to train
our model on datasets of fruits and vegetables. We used
FIGURE 5 | Gas detection sensor.
the Fruits360 dataset, which is openly available and may be
downloaded from GitHub or found on Kaggle (41). It has
around 90,380 pictures in various categories. This dataset
contains high-resolution pictures, which are required for a
successful classifier.
In our model, there are a total of 11 layers. The output
layer is SoftMax, and there are four convolutional layers,
four max pooling layers, and two fully connected layers. The
input layer is a convolutional layer with a 16, 5 x 5 x 4
filter and 2 x 2 max-pooling with stride = 2. Convolutional
layer 4 has a parameter of (5 x 5 x 64) and a dimension
of 128. The model has been trained on 50 different types
of fruits and vegetables, and it is also capable of identifying
multiclass images.

7. RESULTS AND DISCUSSION

Information gathering is an important and time-consuming
FIGURE 6 | Humidity sensor. task. Regardless of the field of study, precise data collection

Frontiers in Public Health | [Link] 8 January 2022 | Volume 9 | Article 816226

Sonwani et al. Food Spoilage Detection and Analysis

FIGURE 7 | Heat sensor and cooling module.

TABLE 3 | Description of the fruits and veggies dataset.

Columns Description Value Type

Names of fruits and vegetables Different types of fruits and vegetables NA String
Minimum optimal storage temperature Minimum temperature in which fruit or Multiple minimum optimal temperature values Numeric
vegetable remain fresh
Maximum optimal storage temperature Maximum temperature in which fruit or Multiple maximum optimal temperature values Numeric
vegetable remain fresh
Freezing point This cooling point in which fruit or vegetable Multiple freezing point values Numeric
remain fresh
Minimum optimal humidity Minimum humidity in which fruit or vegetable Multiple minimum optimal humidity values Numeric
remain fresh
Maximum optimal humidity Maximum humidity in which fruit or vegetable Multiple maximum optimal humidity values Numeric
remain fresh
Minimum approximate storage life At least number of days in which fruit or Multiple minimum approximate storage life values Numeric
vegetable remain fresh
Maximum approximate storage life At most number of days in which fruit or Multiple maximum approximate storage life values Numeric
vegetable remain fresh
Average shelf life Average of minimum (start spoiling) spoilage Multiple average shelf life values Numeric
time and maximum (after spoiled) spoilage time

is crucial to maintaining cohesion. Food spoilage is a very We have used this dataset for CNN model training and
concerning topic nowadays because this time is needed to object detection.
have a strong immune system, and there is no existing Fruits and Veggies—We have created a dataset after careful
solution that helps to prevent spoilage to some extent. The research. The dataset contains 50 types of fruits and vegetables
Fruit360 is the data set that was used in this study. We with their optimal storage temperature range, optimal humidity
have to gather a dataset from the Kaggle dataset website. range, approximate storage life range, and freezing point are
The dataset contains images of different fruits and vegetables. included in this collection (42). Dataset is used for spoilage alert

Frontiers in Public Health | [Link] 9 January 2022 | Volume 9 | Article 816226

Sonwani et al. Food Spoilage Detection and Analysis

FIGURE 8 | Comparison among different fruits and vegetables with respect to their various attributes.

generation messages in mobile. Table 3 describes the dataset The performance graph for our model is plotted for both the
attribute with its value, type, and description. All the value of each training and validation data. We show the training data loss vs.
column represents numeric and floating point values except the validation data loss graph, as well as the training data accuracy
first column. and validation data accuracy of the CNN model for fruits and
Figure 8 represents the relation between attributes like vegetables, for object detection. The performance graph plot
minimum optimal storage temperature, maximum optimal shows the improvement in accuracy as the number of epochs
storage temperature, Freezing Point, minimum optimal increases, as well as whether or not our model has been correctly
humidity, maximum optimal humidity, minimum approximate trained. The CNN model’s training data loss vs. validation
storage life, and maximum approximate storage life of 17 fruits data loss, as well as training data accuracy and validation data
and vegetables. We have selected 15 fruits or vegetables out of accuracy, are shown on the graph shown in Figures 9, 10. There
50 fruits or vegetables for our experimental purpose. This figure are three different fit situations for the model: over fitting, under
represents the comparison among standard parameter settings of fitting, or perfect fit.
different food items. Then we put the food item in the proposed
device, which further analyzes the item and sets the parameter • Over fitting: When the Training loss is considerably less than
values accordingly. Finally, the proposed approach is used to the Validation loss, this scenario occurs.
monitor the item and maintain the environment accordingly. • Under fitting: This situation occurs when the Training loss
In the end, we monitored the final shelf life of the food item. In exceeds the Validation loss, i.e., Training loss > Validation loss.
this experiment, we try to extend the maximum approximate • Perfect fit: This circumstance occurs when Training
storage life of fruits and vegetables, which is represented and Validation losses are approximately equal or
in Table 4. converging over time, indicating that we are doing
The average shelf life of any fruit or vegetable is a time period things correctly.
in which fruits or vegetables are healthy for eating. After this As seen in our graph of Figure 9, it is virtually converging to
period of time, it is harmful for health to consume this fruit indicate that our model is free of over fitting and under fitting.
or vegetable as spoilage may have taken place previously before The plot of training and validation accuracy after each epoch
the average shelf life. Every fruits or vegetables have a different is shown in Figure 10, our model is virtually ideal in this case
minimum optimal storage temperature below which they started since the validation and training accuracy are nearly identical. So,
spoiling. So, careful minimum optimal storage temperature must our model has a 95% accuracy rate, which is excellent enough to
be maintained. recognize the kind of vegetable and fruit from a picture. It also

Frontiers in Public Health | [Link] 10 January 2022 | Volume 9 | Article 816226

Sonwani et al. Food Spoilage Detection and Analysis

TABLE 4 | Experimental analysis of fruits and vegetables.

Name of fruits or vegetables Minimum Maximum Average shelf Maximum approximate After experimental
temperature (◦ F) temperature (◦ F) life (days) storage life (days) analysis (days)

Broccoli 32 32 11 14 16
Cabbage (Early) 32 32 41 42 44
Carrots (Immature) 32 32 35 180 181
Cauliflower 32 32 14 120 122
Cherries 30 31 6 14 15
Grapes 31 32 6 56 55
Kohlrabi 32 32 7 90 91
Gooseberries 31 32 3 28 29
Leeks 32 32 11 90 91
Parsley 32 32 6 90 91
Plums 31 32 4 35 36
Eggplant 46 54 2 7 9
Blackberries 32 33 6 3 4
Corn (Sweet) 32 32 7 8 9
Cucumbers 50 55 11 14 15

FIGURE 9 | Performance graph of Training Loss vs. Validation Loss.

FIGURE 10 | Performance graph of training accuracy vs. validation accuracy.

Frontiers in Public Health | [Link] 11 January 2022 | Volume 9 | Article 816226

Sonwani et al. Food Spoilage Detection and Analysis

works well for single items, such as single fruits or vegetables. It DATA AVAILABILITY STATEMENT
also has the ability to forecast the class of numerous objects. For
50 different sorts of categories, the approach works perfectly. It The datasets presented in this study can be found in online
provides extremely precise results for the photos that are fed into repositories. The names of the repository/repositories and
the model. accession number(s) can be found below: [Link]
com/spreadsheets/d/1Cz8N7z6tJ8FouYnbbn5shy2D71Ln5P0W/
edit?usp=sharing&ouid=105439467837665352046&rtpof=true&
8. CONCLUSION AND FUTURE SCOPE sd=true.

This study presents a novel technique for Monitoring and

analysis of food spoilage using a sensor bases system. The device AUTHOR CONTRIBUTIONS
proposed in this study is able to preserve food for more days.
Additionally, food items can be prevented from getting spoiled by UB and ES carried out the experiment. ES wrote the manuscript
increasing their lifespan. It monitors the quality of food items and with support from UB. ES fabricated the device with the
keeps notifying the user with voice-activated commands or via help from RA, AB, and MH. UB supervised the project.
display, and it also generates alerts to the user with the predicted All authors contributed to the article and approved the
remaining time of the food spoilage. The proposed device shows submitted version.
an accuracy of 95%.
The proposed smart device can be improved by applying
image processing and machine learning algorithms to detect ACKNOWLEDGMENTS
early spoilage. This can be utilized in refrigeration systems for
detecting food items, spoilage and monitoring for prevention The authors are grateful to the Taif University Researchers
of food spoilage. The device can be incorporated into food Supporting Project (Number TURSP-2020/36), Taif University,
transportation containers which would allow tracking and Taif, Saudi Arabia and Dr B. R. Ambedkar National Institute
detecting the spoilage if any during transportation. The device of Technology Jalandhar for their support to carry out
could also be tested for different variety of foods as well. this work.

REFERENCES 9. Brody AL, Bugusu B, Han JH, Sand CK, McHugh TH. Innovative
food packaging solutions. J Food Sci. (2008) 73:107–116.
1. Dhanamjayulu C, Nizhal U, Maddikunta PKR, Gadekallu TR, Iwendi C, Wei doi: 10.1111/j.1750-3841.2008.00933.x
C, et al. Identification of malnutrition and prediction of BMI from facial 10. Pang Z, Chen J, Zhang Z, Chen Q, Zheng L. Global fresh food
images using real-time image processing and machine learning. IET Image tracking service enabled by wide area wireless sensor network. In:
Proc. (2021). 1–12. doi: 10.1049/ipr2.12222 2010 IEEE Sensors Applications Symposium (SAS). Limerick: IEEE (2010).
2. Ahmad P, Jin H, Alroobaea R, Qamar S, Zheng R, Alnajjar F, et al. p. 6–9.
MH UNet: a multi-scale hierarchical based architecture for medical image 11. Drago E, Campardelli R, Pettinato M, Perego P. Innovations in smart
segmentation. IEEE Access. (2021) 9:148384–408. doi: 10.1109/ACCESS.2021. packaging concepts for food: an extensive review. Foods. (2020) 9:1628.
3122543 doi: 10.3390/foods9111628
3. Arul R, Alroobaea R, Mechti S, Rubaiee S, Andejany M, Tariq U, et al. 12. Bourouis S, Alroobaea R, Rubaiee S, Andejany M, Almansour FM, Bouguila
Intelligent data analytics in energy optimization for the internet of underwater N. Markov chain monte carlo-based bayesian inference for learning finite
things. Soft Comput. (2021) 25:12507–19. doi: 10.1007/s00500-021- and infinite inverted beta-liouville mixture models. IEEE Access. (2021)
06002-x 9:71170–83. doi: 10.1109/ACCESS.2021.3078670
4. Anajemba JH, Tang Y, Iwendi C, Ohwoekevwo A, Srivastava G, Jo O. Realizing 13. Iwendi C, Khan S, Anajemba JH, Bashir AK, Noor F. Realizing
efficient security and privacy in IoT networks. Sensors. (2020) 20:2609. an efficient IoMT-assisted patient diet recommendation system
doi: 10.3390/s20092609 through machine learning model. IEEE Access. (2020) 8:28462–74.
5. Gadekallu TR, Khare N, Bhattacharya S, Singh S, Maddikunta PKR, Ra doi: 10.1109/ACCESS.2020.2968537
IH, et al. Early detection of diabetic retinopathy using PCA-firefly based 14. Hebbar N. Freshness of food detection using IoT and machine
deep learning model. Electronics. (2020) 9:274. doi: 10.3390/electronics learning. In: 2020 International Conference on Emerging Trends in
9020274 Information Technology and Engineering (ic-ETITE). Vellore: IEEE (2020).
6. Gadekallu TR, Khare N, Bhattacharya S, Singh S, Maddikunta PKR, p. 1–3.
Srivastava G. Deep neural networks to predict diabetic retinopathy. J 15. Kumar A. An arduino sensor-based approach for detecting the food spoilage.
Ambient Intell Humanized Comput. (2020) 1–14. doi: 10.1007/s12652-020-01 Int J Eng Appl Sci Technol. (2020) 5:596–9.
963-7 16. Chander BR, Kumari G. Food quality monitoring system by using arduino. J
7. Ayoub A, Mahboob K, Javed AR, Rizwan M, Gadekallu TR, Abidi MH, Eng Sci. (2020). 11:625–30.
et al. Classification and categorization of covid-19 outbreak in pakistan. 17. Ishangulyyev R, Kim S, Lee SH. Understanding food loss and waste–Why
Comput Mater Continua. (2021). 69:1253–69. doi: 10.32604/cmc.2021. are we losing and wasting food? Foods. (2019) 8:297. doi: 10.3390/foods
015655 8080297
8. Jedermann R, Nicometo M, Uysal I, Lang W. Reducing Food 18. Yousefi H, Ali MM, Su HM, Filipe CD, Didar TF. Sentinel wraps: real-time
Losses by Intelligent Food Logistics. London: The Royal Society monitoring of food contamination by printing DNAzyme probes on food
Publishing (2014). packaging. ACS nano. (2018) 12:3287–3294. doi: 10.1021/acsnano.7b08010

Frontiers in Public Health | [Link] 12 January 2022 | Volume 9 | Article 816226

Sonwani et al. Food Spoilage Detection and Analysis

19. Green GC, Chan AD, Goubran RA. Tracking food spoilage in the FEMS Microbiol Lett. (2008) 281:1–9. doi: 10.1111/j.1574-6968.2008.
smart home using odour monitoring. In: 2011 IEEE International 01084.x
Symposium on Medical Measurements and Applications. Bari: IEEE (2011). 33. Raveau R, Fontaine J, Lounès-Hadj Sahraoui A. Essential oils as potential
p. 284–7. alternative biocontrol products against plant pathogens and weeds: a review.
20. Megalingam RK, Sree GS, Reddy GM, Krishna IRS, Suriya L. food spoilage Foods. (2020) 9:365. doi: 10.3390/foods9030365
detection using convolutional neural networks and k means clustering. 34. Bastos DH, Saldanha LA, Catharino RR, Sawaya A, Cunha IB, Carvalho
In: 2019 3rd International Conference on Recent Developments in Control, PO, et al. Phenolic antioxidants identified by ESI-MS from yerba mate (Ilex
Automation & Power Engineering (RDCAPE). Noida: IEEE (2019). p. 488–93. paraguariensis) and green tea (Camelia sinensis) extracts. Molecules. (2007)
21. Green GC, Chan AD, Goubran RA. Monitoring of food spoilage with 12:423–32. doi: 10.3390/12030423
electronic nose: potential applications for smart homes. In: 20093rd 35. Anand G, Prakash L. IoT based novel smart refrigerator to curb food wastage.
International Conference on Pervasive Computing Technologies for Healthcare. In: 2018 3rd International Conference on Contemporary Computing and
London: IEEE (2009). p. 1–7. Informatics (IC3I). Gurgaon: IEEE (2018). p. 268–72.
22. Yuan M, Ghannam R, Karadimas P, Heidari H. Flexible RFID patch for food 36. Chanthini B, Manivannan D, Umamakeswari A. Perishable food quality
spoilage monitoring. In: 2018 IEEE Asia Pacific Conference on Postgraduate monitoring–an internet of things (IoT) approach. Int J Pure Appl Math. (2017)
Research in Microelectronics and Electronics (PrimeAsia). Chengdu: IEEE 115:63–66. Available online at: [Link]
(2018). p. 68–71. 7/[Link]
23. Caya MVC, Cruz FRG, Blas PJR, Cagalingan MM, Malbas RGL, Chung WY. 37. Ilie AM, Vaccaro C. Design of a smart gas detection system in areas
Determining spoilage level against time and temperature of tomato-based of natural gas storage. In: 2017 IEEE International Geoscience and
Filipino cuisines using electronic nose. In: 2017IEEE 9th International Remote Sensing Symposium (IGARSS). Fort Worth, TX: IEEE (2017).
Conference on Humanoid, Nanotechnology, Information Technology, p. 5954–7.
Communication and Control, Environment and Management (HNICEM). 38. Jain P, Kushwaha R. Wireless gas sensor network for detection and monitoring
Manila: IEEE (2017). p. 1–5. of harmful gases in utility areas and industries. In: 2012 Sixth International
24. Ponnan S, Saravanan AK, Iwendi C, Ibeke E, Srivastava G. An Conference on Sensing Technology (ICST). Kolkata: IEEE (2012). p. 642–6.
artificial intelligence based quorum system for the improvement of 39. Jovanović U, Jovanović I, Mančić D. Overview of temperature sensors for
the lifespan of sensor networks. IEEE Sens J. (2021) 21:17373–85. temperature measurement of PV modules. In: 2018 26th Telecommunications
doi: 10.1109/JSEN.2021.3080217 Forum (TELFOR). Belgrade: IEEE (2018). p. 1–8.
25. Jorge-Martinez D, Butt SA, Onyema EM, Chakraborty C, Shaheen Q, De- 40. Benabdellah N, Bourhaleb M, Benazzi N, Dahbi S. Design of temperature
La-Hoz-Franco E, et al. Artificial intelligence-based Kubernetes container for and humidity sensors for an electronic nose used in rotten food. In: 2016
scheduling nodes of energy composition. Int J Syst Assurance Eng Manag. International Conference on Electrical and Information Technologies (ICEIT).
(2021). 1–9. doi: 10.1007/s13198-021-01195-8 Tangiers: IEEE (2016). p. 505–9.
26. Saggin B, Belaizi Y, Vena A, Sorli B, Guillard V, Dedieu I. A Flexible 41. Oltean M. Fruit-360 Dataset Available. (2021). Available online at: https://
Biopolymer based UHF RFID-Sensor for food quality monitoring. In: 2019 [Link]/moltean/fruits
IEEE International Conference on RFID Technology and Applications (RFID- 42. Fruit and Veggies Dataset Available. (2021). Available online at: [Link]
TA). Pisa: IEEE (2019). p. 484–7. [Link]/spreadsheets/d/1Cz8N7z6tJ8FouYnbbn5shy2D71Ln5P0W/edit?
27. Law SE, Cooper SC. Air-assisted electrostatic sprays for postharvest control usp=sharing&ouid=105439467837665352046&rtpof=true&sd=true
of fruit and vegetable spoilage microorganisms. IEEE Trans Ind Appl. (2001)
37:1597–602. doi: 10.1109/28.968166 Conflict of Interest: The authors declare that the research was conducted in the
28. Phukkaphan N, Eamsa-ard T, Chairanit C, Kerdcharoen T. The application absence of any commercial or financial relationships that could be construed as a
of gas sensor array based electronic nose for milk spoilage detection. In: 2021 potential conflict of interest.
7th International Conference on Engineering, Applied Sciences and Technology
(ICEAST). Pattaya: IEEE (2021). p. 273–6. Publisher’s Note: All claims expressed in this article are solely those of the authors
29. Goel P, Goel S, Bhatia S. Food quality assessment using fuzzy and do not necessarily represent those of their affiliated organizations, or those of
logic. In: 2015 2nd International Conference on Computing for
the publisher, the editors and the reviewers. Any product that may be evaluated in
Sustainable Global Development (INDIACom). New Delhi: IEEE (2015).
this article, or claim that may be made by its manufacturer, is not guaranteed or
p. 1459–62.
30. Yu B, Zhan P, Lei M, Zhou F, Wang P. Food quality monitoring endorsed by the publisher.
system based on smart contracts and evaluation models.
IEEE Access. (2020) 8:12479–90. doi: 10.1109/ACCESS.2020.296 Copyright © 2022 Sonwani, Bansal, Alroobaea, Baqasah and Hedabou. This is an
6020 open-access article distributed under the terms of the Creative Commons Attribution
31. Prajwal A, Vaishali P, Sumit D. Food quality detection and monitoring system. License (CC BY). The use, distribution or reproduction in other forums is permitted,
In: 2020 IEEE International Students’ Conference on Electrical, Electronics and provided the original author(s) and the copyright owner(s) are credited and that the
Computer Science (SCEECS). Bhopal: IEEE (2020). p. 1–4. original publication in this journal is cited, in accordance with accepted academic
32. Considine KM, Kelly AL, Fitzgerald GF, Hill C, Sleator RD. High- practice. No use, distribution or reproduction is permitted which does not comply
pressure processing-effects on microbial food safety and food quality. with these terms.

Frontiers in Public Health | [Link] 13 January 2022 | Volume 9 | Article 816226