Insect and Pest Detection in Stored Grains: Analysis of Environmental
Factors and Comparison of Deep Learning Methods
DEVI PRIYA R., ANITHA N., DEVISURYA V., VIDHYAA V. P., SHOBIYA K., SUGUNA C.
Department of Information Technology,
Kongu Engineering College,
Erode, Tamil Nadu,
INDIA
Abstract: Majority of the world’s population depends on agro-based economy for their income and survival. In
developing and under-developed countries, due to reasons like basic farming techniques, less educational and
technological exposure, lack of technological advancements and recent agricultural knowledge, yield of the crops is
very low and moreover there is a huge loss during storage also. Insects, pests and diseases more often affect the
stored grains and cause heavy damage to the quantity and quality of the grains. Insecticides and pesticides cannot
provide better solution all the times and hence there is an acute need for computer vision based techniques capable
of monitoring the spread of insects in the initial stages of storage and protecting the stored grains from further
damages and losses. Hence, this paper provides analysis of various factors which can cause damage to the stored
grains natural ways to protect crops. It provides the comparison results of various standard deep learning methods
that are used to detect the insects and pests in stored grains.
Keywords: Insects and pests, stored grains, CNN, environmental factors, YOLOv5 algorithm.
Received: June 27, 2021. Revised: May 2, 2022. Accepted: May 25, 2022. Published: June 15, 2022.
1 Introduction
Agriculture serves as the main occupation for people
all over the world since it is the main source of
livelihood. In most of the developing countries,
around 70% of the population depend on agriculture
for their source of income. Agricultural productivity
and consumption is affected by various
environmental and other human factors [1,2,3]. Even
though, globally steady improvement can be seen in
grain production due to advancements in technology,
post harvest loss is estimated at around 25 % each
year [4]. Out of the total loss faced, 6% of the loss is
due to loss of grains during storage after harvesting.
Grain storage is a critical phase in order to get
maximum profit in agricultural domain in which loss
may occur due to invasion of insects, pests,
pathogens and rodents. The insects and pests still
reduce the quality of remaining stored grains. Out of
the total loss during grain storage, technical
inefficiency accounts to 50% of the loss
approximately.
The common grain pests namely lesser grain
borer, rice weevil and rust red flour beetle increase
loss of grains and their management difficulty in two
important ways: i). Directly impact the management
cost for the farmers through the expense of pest
control in the farms and ii). Increase the cost for pest
management for grain storage authorities in bulk
storage areas. Grain insect pests can be classified into
primary and secondary pests. The primary grain
insects affect only the fresh and whole unbroken
grains whereas the secondary pests feed on already
damaged grain, dust and milled products during
storage.
Different losses that are faced commonly while
storing grains are given below:
i. Quantitative loss: If the insects feed directly into
the grains, it can cause heavy loss in weight of the
grains. For example, common pest namely rice
weevil consumes about 14g from 20 mg rice for
its development. The total weight of the grains is
lost to a maximum extent and cause heavy loss to
the farmers.
ii. Qualitative loss: Most of the grain pests consume
grain embryos which in turn lower the protein
content of the grains and also the quality of seeds
capable of germination. The chemical components
WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT
DOI: 10.37394/232015.2022.18.71
Devi Priya R., Anitha N.,
Devisurya V., Vidhyaa V. P., Shobiya K., Suguna C.
E-ISSN: 2224-3496
759
Volume 18, 2022
of the grain will get affected and the contaminated
grain with infected skin and body parts further
attract the spread of pathogenic microorganisms.
iii. Loss of seed viability: The viability of the seeds
will be severely affected and cannot be used for
sowing and further plantation. The capability of
seeds to develop into plants will get degraded.
iv. Damage of storage containers: The grains are
usually stored in wooden containers, polythene,
lined bags, sacks, etc. Some pests like lesser grain
borer can damage these containers and hence the
grains will be lost and wasted to a great extent. If
the storage container is damaged, people will have
the tendency to avoid buying those products out
of fear about its quality.
Small farmers use to store their grains in small
amount in their house itself whereas large amount of
grains like rice, wheat, turmeric, millets are stored in
warehouses for use in the required time period. In
spite of the artificial pesticides and insecticides to
protect them, considerable amount of loss cannot be
avoided. In order to achieve maximum yield, the
agricultural processes need to be integrated with
modern technological interventions equipped with
artificial intelligence techniques like machine
learning and deep learning methods. It can help in
eliminating the harmful insects that damage grains
and hence improve the productivity of crops.
2 Literature Review
The methods which are commonly used for insect
and pest detection and controlling of environmental
parameters to avoid invasion of insects and pests in
the stored grains are discussed below.
During the earlier days, people have used trap
types for catching pests and insects. Different types
of traps were placed near the location where grains
are stored. If any insects come near this location, they
will be stick and caught in the trap. But, that will be
manual procedure and is very inefficient. Traditional
methods of pest management like near infrared,
acoustic methods and electrical conductivity create
lot of difficulties in sampling, reduction in speed and
lot of manual workload. Initially, machine learning
based image recognition methods are commonly used
by researchers for a long time. Deep learning
methods are gaining popularity in the recent days by
successful implementation and better classification
results in different applications [5].
In [6], authors have insisted that many
researchers were using image analysis techniques that
automatically scan X-ray images to detect insect
infestations. The major problems in X-ray and NIR
spectroscopy methods are that they are very costly
and needs complex operating mechanisms which are
very difficult for a farmer with low technical
expertise. In [7], spatial association maps are used
which have positive value if there are insects and
mites inside the bin in which grains are stored and
negative value if not. The association patterns
between two adjacent samplings are analyzed and if
there are no insects and mites, the association will be
higher. If there are some trapped insects and mites,
the association value might become low. Various
statistical methodologies that are being used for in-
storage sampling and surveillance in the grains
warehouse are discussed in [8].
[9] has assessed, evaluated and critically
analyzed the techniques used for judicious pest
management in food storage. It presents and analyzes
a variety of methods in real world applications. [10]
has discussed about Integrated pest management
(IPM) and the recent methods that are used for IPM
are briefly reviewed. [11] made decisions on
controlling pests using population dynamics and
threshold insect densities. They insist that better
sampling methods are very much essential for
securing postharvest food with the sharp increase in
human population. [12] has used radio-frequency
grain bin imaging system to monitor and control
stored grains. The dielectric properties of the grain
are monitored to check moisture content,
temperature, insect invasion and other abnormal
changes. The captured values in images are processed
to serve the purpose.
In [13], Multispectral Imaging (MSI) technology
is combined with chemometrics to identify the
variations between intact and insect-infested
almonds. Principal Component Analysis (PCA) and
Support Vector Machines (SVM) are used to classify
them with better prediction results of 97%. [14] has
researched on occurrence of stored grain pests in the
underground pit grain storages of Eastern Ethiopia
and found that around 70% of the grains were
infected with mean germination in at least 7-8
months period. [15] has developed a method for
detecting and classifying six different insects in the
stored grain. RGB images of the live insects were
used and Faster R-CNN based improved inception
network is used to extract feature maps.
WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT
DOI: 10.37394/232015.2022.18.71
Devi Priya R., Anitha N.,
Devisurya V., Vidhyaa V. P., Shobiya K., Suguna C.
E-ISSN: 2224-3496
760
Volume 18, 2022
Deep learning methods have gained significance
in recent years in food sensory and consuming
researches. [16] has provided a detailed review of
different deep neural network algorithms in food
industry for ensuring food quality and safety
inspection measures. They have found that deep
learning algorithms outperform other machine
learning algorithms for feature selection and further
data analysis. [17] has proposed implementing
acoustic technology with visual surveys and pitfall
traps to identify and detect insects in Kenyan data
warehouses to prevent wastage in postharvest
maintenance. The measures are also taken to identify
background noise. It is very much essential to reduce
the losses incurred in postharvest maintenance phase.
[18] has used state-of-the-art IoT enabled system
to monitor temperature, relative humidity and carbon
dioxide and predict the type of insect activity in
stored grains. DHT22 and CDM7160 sensors were
used for this purpose. Only the abstract details of
whether it is infected or not can be identified using
this method and detailed analysis of the level of
infection and controlling mechanisms are not
possible in this approach. [19] has devised a portable
postharvest insect detection system where electret
microphones are used to record insect sounds. The
sounds of insects captured are analyzed by custom
written software which compares them with sounds
of known pets. The sounds of 5 to 50 insects are
differentiated by aggregation pheromones or other
active semio chemicals.
[20] has categorized pesticides into four different
forms namely gas, liquid, gel/foam and solid.
Conventional strategies include usage of insecticide
baits, aerosols, sprays, fumigants and inert gases.
Food protection under postharvest condition may
improve if these methods are improved or hybridized
with other methods. Electrostatic dusts or sprays,
nanoparticles, hydrogels, inert baits with artificial
attractants, biodegradable cyanogenic protective
grain coating are some of the advanced technologies.
In [21], acoustic detection of immature insects hidden
within the stored grains is proposed. The immature
insects are large in number and are often present
without adult insects inside the grains. Modern
acoustic tools can effectively detect insects based on
their images.
3 Environmental Factors Affecting
Insects and Pests Management
Rather than implementing various machine learning
and deep learning methods for detecting the insects
and pests in the stored grains, it is always better to
prevent the invasion of insects and pests in the
beginning itself. This section discusses about various
environmental factors which influence the attack of
insects and pests and highlights natural ways to
protect the grains without being attacked.
3.1 Environmental Parameters
Insects and pests living in the stored grains rely on
their food (grain) for the water required for its
survival and hence there is no necessity of external
water source. In general, if the moisture content of
the grain is low (less than 10%), insects either tend to
break the stored grains or utilize its own water energy
stored with fatty tissues. By doing so, some of the
insects survive and the remaining insects will not be
able to survive and increase its population. In
general, when the moisture content of the grains and
the storage containers is above 15%, there is a chance
for rapid increase in population of the insects. Studies
have shown that some species of insects like foreign
grain beetle and the larger black four beetle populate
at a high speed in high moisture when compared with
low moisture regions.
Most of the stored grain insects develop within
short period of time at room temperature itself with
high reproduction rate where some female insects can
lay down 100-400 eggs. The life span of adult insects
also ranges from weeks to years. Two primary
environmental factors which impact the growth of
insects are temperature and moisture level. The
insects usually require temperature of around 15-
20°C for survival and reproduction and at around
25°C, its population start damaging the crops. When
the temperature is increased beyond 35°C, the insect
pests cannot survive and stop laying eggs and hence
they will vanish soon. But, the challenging factor is
that grain is an excellent insulator and hence the
required air supply is provided to the insects during
severe cold or winter seasons. Sometimes, when the
temperature is too low for dyeing, some species still
survive but they cannot feed and die out of starvation.
Molds may grow in stored grains when the moisture
content is greater than 14.5%. In some cases of
species, the molds can produce mycotoxins which are
harmful for health factors and seriously affect agro-
WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT
DOI: 10.37394/232015.2022.18.71
Devi Priya R., Anitha N.,
Devisurya V., Vidhyaa V. P., Shobiya K., Suguna C.
E-ISSN: 2224-3496
761
Volume 18, 2022
food industries. More suitable temperature for grain
molds is 25-30°C. Some species like Aspergillus spp
grow well even at high temperature above 35°C.
Spores of storage fungi usually occur during the
period of harvesting, transporting and handling
procedures. While storing, when the temperature and
moisture levels are suitable, fungi start germinating
and its growth is unstoppable.
3.2 Environmental Solutions to Control
Invasion of Insects and Pests in Stored Grains
In order to avoid insect pest invasion into the stored
grains, ventilator facility for required air exhaust is
recommended. This can be done artificially by either
positive pressure or negative pressure. The storage
structures need to be equipped with provision of
sucking the moisture or air out of the stored grains. In
some environments, the temperature may not be
sufficient to cool the grain. Refrigerated air may be
supplied sometimes to meet the requirements. But,
this method may look so expensive to implement for
all the cases. However, this method can be used for
some expensive grains. The refrigeration unit will be
connected by an insulating pipe for protecting the
grains from insect pests by grain chilling. In sub-
tropical countries, partial dried grains are processed
with dryeration and batch drying. In regions with
warm climate, the grains are immediately stored in
order to preserve the germination capability of the
grains.
Artificial pesticides are often found to be effective in
controlling the pests but cannot be used continuously
by all categories of farmers due to its non-
biodegradability, high cost and the negative impacts
on human and also soil health. Hence, most of the
agricultural practitioners are seeking alternative
powerful and eco-friendly natural ways of pest
control with low cost. Plant volatile organic
compounds are being increasingly used to protect the
grains against insects and pests. A broad review of
plant volatile organic compounds commonly used to
protect grains is given in [22].
4 Comparison of Deep Learning
Methods for Detection
The deep learning methods namely Convolution
Neural Network (CNN), Fast Region-based CNN
(Fast R-CNN), Faster Region-based CNN (Faster
R- CNN) and You Look Only Once v5
(YOLOv5) which are used to detect the presence
of insects and pests in the stored grains are
briefly described below.
4.1 CNN
The Convolutional neural network is a standard deep
learning algorithm used for image processing which
contains convolution, pooling and fully connected
layers as given in fig.1. Multiple stacks of these
layers can be constructed in order to achieve better
learning and classification. The image is convolved
with filters also known as kernel. If the input image
is given by X and filter by f, the output of
convolution operation (*) represented by Z is given
as Z=X * f (1)
In the convolution layers, significant feature maps are
extracted which contains Rectified Linear Unit
(ReLu) as the activation function. For the input x,
RELU function is calculated as
RELU (x) =
(2)
If the dimension of input is (n, n) and that of filter is
(f, f), then
Dimension of output = ((n-f+1), (n-f+1)) (3)
Flatten layer then flattens the features maps obtained
from convolution layer and passes them to fully
connected layers. Fully connected layer implements
linear and non-linear transformation operations. The
execution on linear transformation is given by
Z=WT.X +b (4)
where W is the weight and b is the bias (constant).
For multiple class labels, softmax activation function
can be used which classifies the input and outputs the
resultant class label for the corresponding input
whose formula is given below.
(5)
where - softmax function, - input vector, –
standard exponential function for and -
standard exponential function for output
WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT
DOI: 10.37394/232015.2022.18.71
Devi Priya R., Anitha N.,
Devisurya V., Vidhyaa V. P., Shobiya K., Suguna C.
E-ISSN: 2224-3496
762
Volume 18, 2022
Fig. 1: CNN Architecture
Fig. 2: Fast R-CNN
Fig. 3: Faster R- CNN
WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT
DOI: 10.37394/232015.2022.18.71
Devi Priya R., Anitha N.,
Devisurya V., Vidhyaa V. P., Shobiya K., Suguna C.
E-ISSN: 2224-3496
763
Volume 18, 2022
Fig. 4: YOLO v5 algorithm
vector. It gives probabilities for every ROI over
(K+1) class labels p = p0, p1, p2,.. pk. The
classification loss
Lcls (p,u) = -log(pu) (6)
4.2 Fast R-CNN
In Fast R-CNN, the operation is fast since the
convolution operation is done only once per image
and the feature map is generated from the output of
convolution operation. Region based ROI pooling is
done and the architecture is given in fig.2. R-CNN
process 2000 region proposals per image and that
overburden is avoided in Fast R-CNN and hence
faster execution can be observed.
4.3 Faster R-CNN
In Faster R-CNN, the image is first provided as input
to the backbone network that generates the
convolution feature map which is then passed to the
Region Proposal Network (RPN). As given in fig.3,
for each sliding window, a maximum of k anchor
boxes are generated and for mini-batches are
generated from them. It receives the feature map and
generates anchors which are given into the
classification layer for classifying the objects with
the resulting bounding box. The training loss for
RPN for multiple classes is given by
(7)
where pi – predicted probability that anchors has
object or not, - ground truth value of anchors has
object or not, and - coordinates of anchors
predicted and ground truth coordinate of bounding
boxes respectively, - classifier loss, –
regression loss, and – normalization
parameters of mini batch size and regression
respectively and – constant.
4.4 YOLO v5 Algorithm
Fig. 4 shows the architecture of YOLOv5 algorithm.
It is an efficient algorithm which is quicker in
process and there is no need of looking at the training
set every time the algorithm iterates. It is looked up
only once and the entire feature maps are studied.
The backbone network can be used with neck and
then followed by dense prediction and exact sparse
prediction following it.
5 Experimental Results and Discussion
In order to validate the efficiency of various deep
learning methods to detect insects and pests in stored
grains, methods like CNN, Fast R-CNN, Faster R-
CNN and YOLOv5 algorithms are compared. The
experiments are conducted on the pest images of
public IP102 dataset. It contains 75,000
Table 1. Classification accuracy of deep learning algorithms with different models
Backbone
architecture
CNN
Fast R-CNN
Faster R-CNN
YOLOv5
Aphids
Cicadellidae
Inception V3
87.21
88.14
87.45
92.32
Xception
88.60
89.48
90.27
93.14
VGG19
89.0
90.16
91.35
92.76
WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT
DOI: 10.37394/232015.2022.18.71
Devi Priya R., Anitha N.,
Devisurya V., Vidhyaa V. P., Shobiya K., Suguna C.
E-ISSN: 2224-3496
764
Volume 18, 2022
ResNet 50
89.23
90.74
92.41
93.65
Flea beetles
Inception V3
83.91
85.58
87.04
89.37
Xception
84.20
85.39
87.37
88.31
VGG19
84.91
84.63
88.34
89.67
ResNet 50
85.67
86.52
88.36
90.50
Cicadellidae
Inception V3
86.74
86.71
87.94
93.41
Xception
87.15
88.47
92.05
93.30
VGG19
88.09
88.52
89.46
93.97
ResNet 50
88.31
89.45
90.11
94.28
Flax budworm
Inception V3
77.01
78.94
80.61
85.30
Xception
78.20
79.65
83.28
87.19
VGG19
78.22
81.17
81.29
88.97
ResNet 50
79.09
82.37
81.63
89.13
Red spider mite
Inception V3
79.14
83.63
86.09
89.34
Xception
80.43
81.25
85.47
89.91
VGG19
81.26
84.79
85.03
90.25
ResNet 50
81.24
86.94
83.72
90.80
images of 102 types of insect species. Due to
practical difficulties in implementing all these species
identification, only 5 species namely aphids, flea
beetles, Cicadellidae, flax budworm and red spider
mite are taken for experimental analysis. Whenever
classification is performed, there are more chances
that samples taken for training and testing may go
imbalanced. If the imbalanced samples are chosen for
experiments, it may create overfitting or underfitting
and classification may go biased [23]. In order to
avoid that, equal number of samples (500 images) are
taken from each class label and the experiments are
conducted. The performance of algorithms like CNN,
Fast R-CNN, Faster R-CNN and YOLOv5
algorithms are compared in terms of classification
accuracy, precision and recall measures which are
calculated as follows:
(8)
(9)
(10)
All the algorithms are allowed to run for 25 epochs
each and the average results of 30 runs are reported
below. Table 1 reports the comparison of
classification accuracy of deep learning methods with
various underlying backbone architectures like
Inception V3, Xception, VGG19 and ResNet models.
It is found that standard CNN implementation shows
low accuracy whereas other CNN variations like Fast
R-CNN and Faster R-CNN report comparatively
better accuracy. Among all the methods which are
compared, YOLOv5 reports better results. ResNet
serves as the efficient backbone architecture model
for all deep learning implementations and hence it is
used in further experiments.
Tables 2 and 3 also demonstrate superior
performance of YOLOv5 algorithm than that of other
CNN variations. YOLOv5 algorithm is simple and
efficient and includes augmentation and hence the
required number of samples are introduced for
training.
Table 2. Precision of pest detection algorithms
CNN
Fast
R-
CNN
Faster
R-
CNN
YOLOv5
Aphids
0.88
0.89
0.91
0.93
Flea beetles
0.87
0.84
0.88
0.91
Cicadellidae
0.82
0.87
0.80
0.88
Flax
budworm
0.84
0.81
0.82
0.88
Red spider
mite
0.78
0.81
0.83
0.91
Table 3. Recall of pest detection algorithms
CNN
Fast
R-
CNN
Faster
R-
CNN
YOLOv5
Aphids
0.84
0.79
0.81
0.84
WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT
DOI: 10.37394/232015.2022.18.71
Devi Priya R., Anitha N.,
Devisurya V., Vidhyaa V. P., Shobiya K., Suguna C.
E-ISSN: 2224-3496
765
Volume 18, 2022
Flea beetles
0.77
0.79
0.80
0.77
Cicadellidae
0.81
0.82
0.83
0.81
Flax
budworm
0.82
0.84
0.86
0.82
Red spider
mite
0.83
0.85
0.88
0.83
Table 4. Execution time of algorithms for pest
detection (in seconds)
CNN
Fast
R-
CNN
Faster
R-
CNN
YOLOv5
Aphids
120
102
89
78
Flea beetles
136
113
104
99
Cicadellidae
147
134
117
102
Flax
budworm
138
126
119
97
Red spider
mite
129
103
97
88
Table 4 also reports that in classifying 5 species of
insects, YOLOv5 algorithm completes the
classification process quicker than other methods
since YOLOv5 looks at the training set only once and
does not spend much time in repeated scanning of the
training set.
6 Statistical Results
The experimental results obtained are validated by
using t-test statistic. The performance of YOLOv5
algorithm which shows better results when compared
with other algorithms are compared pairwise using t
test. The t statistic is calculated using the formula
(11)
where X1 and X2 are the mean classification accuracy
of algorithms 1 and 2 which are compared, S1 and S2
represent standard deviation of the algorithms 1 and
2, n1 and n2 indicate the number of observations in
algorithms 1 and 2 respectively. The number of
observations taken from each pair of algorithm is 15.
Only, the top 15 observations with high classification
accuracy are taken into consideration. The alpha
value is taken as 0.01.
Degrees of freedom is 15+15-2 =
28.
The corresponding t value at the given significance
level in the t table is 2.467. From table 5, it is
inferred that the t value obtained by comparing
YOLOv5 with each other algorithm is comparatively
greater than the t table statistic 2.467. Hence, it is
understood that there is a significant difference
between YOLOv5 and other algorithms. Also, among
the compared algorithms, YOLOv5 is close in
performance to Faster R-CNN and hence its t value is
less than that of other algorithms. Higher the t value
obtained, higher the significant difference between
the algorithms that are compared.
Table 5. Comparison of YOLOv5 with other
algorithms using t test
Algorithms compared
Estimated
t statistic
YOLOv5 Vs CNN
5.34
YOLOv5 Vs Fast R- CNN
4.97
YOLOv5 Vs Faster R-CNN
4.02
7 Conclusion
This paper has thus provided a brief overview of
various environmental factors which influence the
growth and spread of insects and pests in the stored
grains and also various natural measures which can
be followed by farmers to control them. In order to
avoid the damages caused due to artificial pesticides
and insecticides, natural and simple method of
implementing image processing algorithm like deep
learning models can be very beneficial. It does not
cause any harm to the nutrient levels of the stored
grains and very cost effective in identification of
insects in the stored grains in the initial stages itself.
It can be very effectively used by small scale and also
large scale farmers and in warehouses to prevent the
invasion of insects into the grains and damages and
in turn economical losses caused by them. Organic
compounds can also be used to protect the grains
from the insects, but it needs extra cost and effort in
implementing them. The better solution can be to
maintain proper environmental parameters like
temperature and moisture contents of the grains,
storage containers and the storage rooms. When
comparing the deep learning methods, YOLOv5
algorithm performs better and that is also validated
by statistical t test results.
WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT
DOI: 10.37394/232015.2022.18.71
Devi Priya R., Anitha N.,
Devisurya V., Vidhyaa V. P., Shobiya K., Suguna C.
E-ISSN: 2224-3496
766
Volume 18, 2022
References:
[1] Yaty S, Zhariani S.M & Yusri S.I. Factors
Influencing Food Consumption among
University Students.WSEAS Transactions on
Business and Economics, (2021), vol. 18, pp.
501-510.
[2] Osabohien R, Olurinola I.O, Matthew O.A,
Igharo A.E. Enabling Environment and
Agriculture in ECOWAS: Implications for
Food Security. WSEAS Transactions on
Environment and Development, (2021),vol. 17,
pp. 38-46,
[3] Almas L.K & Usman M. Determinants of
Wheat Consumption, Irrigated Agriculture, and
Food Security Challenges in Egypt. WSEAS
Transactions on Environment and
Development, (2021), vol. 17, pp. 696-712.
[4] Shankar Uma & Abrol D.P. Integrated Pest
Management in Stored Grains. In: Integrated
Pest Management: principles and Practice,
pp.386-407, (2012).
[5] Shi Z, Dang H, Liu Z, & Zhou X. Detection
and Identification of Stored-Grain Insects
Using Deep Learning: A More Effective
Neural Network. IEEE Access. (2020), 8,
pp.163703 – 163714.
[6] Neethirajan S, Karunakaran C &Jayasb D.S.
White N.D.G. Detection techniques for stored-
product insects in grain. Food Control, (2007),
Vol. 18(2), pp. 157-162.
[7] Athanassiou C.G, KavallieratosN.G,
Sciarretra A, Palyvos N.E, Trematerra P.
Spatial associations of insects and mites in
stored wheat. Journal of Econ Entomology
(2011), 104(5), pp.1752-64.
[8] Elmouttie D, Hammond N.E.B & Hamilton G.
A review of current statistical methodologies
for in-storage sampling and surveillance in the
grains industry. Bulletins of Entomology
Research, (2013), 103(2), pp.140-7.
[9] Athanassiou C.G. & Arthur F.H. Recent
Advances in Stored Product Protection.
Springer; Berlin/Heidelberg, Germany: (2017),
273.
[10] Hagstrum DW, Athanassiou CG. Improving
Stored Product Insect Pest
Management: From Theory to Practice. Insects.
(2019), 4, 10(10):332.
[11] Hagstrum D.W., Phillips T.W. Evolution of
Stored-Product Entomology: Protecting the
World Food Supply. Annu. Rev. Entomol.
(2017), 62, pp.379–397.
[12] Gilmore C, Jeffrey I, Vetri JL & Paliwal J.
Detection and continuous monitoring of
localised high-moisture regions in a full-scale
grain storage bin using electromagnetic
imaging Biosystems Engineering, (2017),
vol.163, pp. 37-49.
[13] Yu, J., Ren, S., Liu, C., Wei, B., Zhang, L.,
Younas, S., & Zheng, L. (2018). Non-
destructive detection and classification of in-
shell insect-infested almonds based on
multispectral imaging technology. The Journal
of Agricultural Science, 156(9), pp.1103-1110.
[14] Tadeos S. Occurrence of Stored Grain Insect
Pests in Traditional Underground Pit Grain
Storages of Eastern Ethiopia. Agri Res & Tech:
Open Access J. (2018),13(2), pp. 555879.
[15] Shen Y, Zhou H, Li J, Jian, F & Jayas, D.
Detection of stored-grain insects using deep
learning. Computers and Electronics in
Agriculture. (2018), 145, pp.319-325.
[16] Zhou L, Zhang C, Liu F, Qiu Z & He Y.
Application of Deep Learning in Food: A
Review. Comprehensive Reviews in Food
Science and Food Safety. (2019), 18, pp.1793-
1811.
[17] Njoroge A, Aognon H, Richter U, Hensel O,
Rohde B, Chen D & Mankin R. Acoustic,
Pitfall Trap, and Visual Surveys of Stored
Product Insect Pests in Kenyan Warehouses.
Insects (2019), 10, 105, pp.1-12.
[18] Sindwani, A. Kumar, C. Gautam, G. Purohit
and P. Tanwar. Prediction and Monitoring of
stored food grains health using IoT Enable
Nodes. IEEE International Conference on
Computing, Power and Communication
Technologies (GUCON), 2020, pp. 516-522.
[19] Mankin RW, Jetter E, Rohde B, Yasir M.
Performance of a Low-Cost Acoustic Insect
Detector System with Sitophilus oryzae
(Coleoptera: Curculionidae) in Stored Grain
and Tribolium castaneum (Coleoptera:
Tenebrionidae) in Flour. J Econ Entomol.
(2020) 9: 113 (6), pp.3004-3010.
[20] Stejskal V, Vendl T, Aulicky R & Athanassiou
C. Synthetic and Natural Insecticides: Gas,
Liquid, Gel and Solid Formulations for Stored-
Product and Food-Industry Pest Control.
Insects, (2021), 12(7): 590.
[21] Mankin, R.; Hagstrum, D.; Guo, M.;
Eliopoulos, P.; Njoroge, A. Automated
WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT
DOI: 10.37394/232015.2022.18.71
Devi Priya R., Anitha N.,
Devisurya V., Vidhyaa V. P., Shobiya K., Suguna C.
E-ISSN: 2224-3496
767
Volume 18, 2022
Applications of Acoustics for Stored Product
Insect Detection, Monitoring, and
Management. Insects, (2021), 12, 259.
[22] DasantaSingh K, Mobolade A.J, Bharali R,
Sahoo D & Rajashekar Y. Main plant volatiles
as stored grain pest management approach: A
review. Journal of Agriculture and Food
Research. (2021), 4, 100127.
[23] R Devi Priya, R Sivaraj, N Anitha, R Rajadevi,
V Devisurya. Variable population sized PSO
for highly imbalanced dataset classification.
Computational Intelligence, vol. 37 (2),
pp.873-890.
Contribution of Individual Authors to the
Creation of a Scientific Article (Ghostwriting
Policy)
Devi Priya R has monitored the project progress and
written the draft of paper
Anitha N and Devisurya V have completed the paper
and performed revisions
Vidhyaa has done analysis of the existing methods
for the chosen problem
Shobiya has designed the experimental setup
Suguna has implemented the algorithms and
performed test analysis
Sources of funding for research presented in a
scientific article or scientific article itself
There are no sources of funding for research
Creative Commons Attribution License 4.0
(Attribution 4.0 International, CC BY 4.0)
This article is published under the terms of the
Creative Commons Attribution License 4.0
https://creativecommons.org/licenses/by/4.0/deed.en_
US
WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT
DOI: 10.37394/232015.2022.18.71
Devi Priya R., Anitha N.,
Devisurya V., Vidhyaa V. P., Shobiya K., Suguna C.
E-ISSN: 2224-3496
768
Volume 18, 2022
