Content uploaded by Gulzar Ahmad
Author content
All content in this area was uploaded by Gulzar Ahmad on Feb 06, 2021
Content may be subject to copyright.
Content uploaded by Muhammad Adnan Khan
Author content
All content in this area was uploaded by Muhammad Adnan Khan on Jan 30, 2021
Content may be subject to copyright.
ech
T
PressScience
Computers, Materials & Continua
DOI:10.32604/cmc.2021.015080
Article
Intelligent Ammunition Detection and Classication System
Using Convolutional Neural Network
Gulzar Ahmad1, Saad Alanazi2, Madallah Alruwaili2, Fahad Ahmad3,6, Muhammad Adnan Khan4,*,
Sagheer Abbas1and Nadia Tabassum5
1School of Computer Science, NCBA&E, Lahore, 54000, Pakistan
2College of Computer and Information Sciences, Jouf University, Sakaka, Aljouf, 72341, Saudi Arabia
3Department of Basic Sciences, Jouf University, Sakaka, Aljouf, 72341, Saudi Arabia
4Riphah School of Computing & Innovation, Faculty of Computing, Riphah International University,
Lahore Campus, Lahore, 54000, Pakistan
5Department of Computer Science, Virtual University of Pakistan, Lahore, 54000, Pakistan
6Department of Computer Sciences, Kinnaird College for Women, Lahore, 54000, Pakistan
*Corresponding Author: Muhammad Adnan Khan. Email: adnan.khan@riphah.edu.pk
Received: 05 November 2020; Accepted: 27 December 2020
Abstract: Security is a signicant issue for everyone due to new and creative
ways to commit cybercrime. The Closed-Circuit Television (CCTV) systems
are being installed in ofces, houses, shopping malls, and on streets to protect
lives. Operators monitor CCTV; however, it is difcult for a single person to
monitor the actions of multiple people at one time. Consequently, there is
a dire need for an automated monitoring system that detects a person with
ammunition or any other harmful material Based on our research and ndings
of this study, we have designed a new Intelligent Ammunition Detection and
Classication (IADC) system using Convolutional Neural Network (CNN).
The proposed system is designed to identify persons carrying weapons and
ammunition using CCTV cameras. When weapons are identied, the cameras
sound an alarm. In the proposed IADC system, CNN was used to detect
rearms and ammunition. The CNN model which is a Deep Learning tech-
nique consists of neural networks, most commonly applied to analyzing visual
imagery has gained popularity for unstructured (images, videos) data classi-
cation. Additionally, thissystem generates an early warning through detection
of ammunition before conditions become critical. Hence the faster and earlier
the prediction, the lower the response time, loses and potential victims. The
proposed IADC system provides better results than earlier published models
like VGGNet, OverFeat-1, OverFeat-2, and OverFeat-3.
Keywords: CCTV; CNN; IADC; deep learning; intelligent ammunition
detection; DnCNN
This work is licensed under a Creative Commons Attribution 4.0 International License,
which permits unrestricted use, distribution, and reproduction in any medium, provided
the original work is properly cited.
2586 CMC, 2021, vol.67, no.2
1 Introduction
In the current era, security has become a vital issue. Under current prevailing conditions of
poor security system, which has caused fear among people, everyone wants to secure his resources;
premises; his organization’s employees and clients. However, CCTV systems are becoming more
popular for providing additional security. They are typically installed in many public places such
as roads, highways, ofces, housing complexes, and shopping malls. Generally, CCTV systems
can see weapons or other harmful items in a person’s hand. However, the potential threat must
be identied by a remote operator, who then may trigger a police response [1]. Moreover, the
images recorded by CCTV might not be examined until after a criminal incident. For this reason,
the system proposed in this study is designed to enable CCTV cameras to detect weapons and
potential threats in real-time.
The primary aim of this research was to minimize or eliminate the threat to security posed
by weapons (e.g., pistols, automatic weapons, and knives) or explosives frequently used in criminal
activities. In this research, we proposed an automated system to detect any dangerous weapons in
the hand of an attacker or a terrorist. This system sounds an alarm that alerts the CCTV oper-
ator, who immediately informs the police or other agencies. It can help citizens before conditions
worsen or the crime is carried out.
The rate of crimes involving weapons is of increasing global concern, specically in those
countries where gun ownership is legal. Initial detection is needed to detect weapons earlier and
allow law enforcement agencies to take immediate action. One of the advanced solutions to this
problem is to supplement surveillance or control CCTV cameras using an automatic system to
detect a pistol, gun, or revolver. It sounds an alarm once it detects any harmful objects in an
attacker’s hand. CNN is also used to detect a rearm using a video recorder [2].
The ammunition detection system includes a rearm and a device for communication. The
rearm contains a transceiver circuit, which detects the release of an ammunition round. The
transceiver circuit produces an electromagnetic signal synchronized with the ammunition discharge.
The communication device, which may be mobile or any handheld device like a radio set, detects
the electromagnetic signal. The modes of communication can be single, half, or full way. The com-
munication device is paired with a geographic location sensor, for example, a GPS receiver [3]. The
communication device contains software that generates an informational message with geographic
information transmitted to a remote device. Later, a converted image is obtained. Another visible
light image, obtained using a transforming system, is included in this transformed image [4]. The
object of interest has, therefore, been identied in the eld [5]. On identication, an indicator
shows the object’s location in the area of interest.
The techniques for monitoring a rearm and generating a warning when the rearm is outside
the designated permit area dened in various scenarios. A location device, such as a Global
Positioning System (GPS) receiver coupled to a transmitter, is connected to or associated with
a rearm. The identied location of the weapon is transmitted to a location service module via
the location device. The position service module compares at least one designated authorized
location with the current rearm location and conrms that the location of the rearm is not
within the authorized permit area. In response, an alert is generated. It is a multi-modal protection
technique that will detect any concealed metallic (weapon or shrapnel), non-metallic (explosives or
Improvised Explosive Device (IED)), or radioactive nuclear threats. Furthermore, long-range facial
recognition of potential terrorists can be carried out by the security checkpoint to detect them.
The security checkpoint integrates many technologies for detecting threats into a single checkpoint
CMC, 2021, vol.67, no.2 2587
designed to be stable and see a wide range of threats, including concealed weapons, explosives,
bombs, and other threats [6].
The importance of security can never be neglected, especially when technological advances
have broadened criminals’ opportunities to commit crimes. After evaluating their security issues,
many corporations have started investing in systems to secure their facilities. New visual systems
that include handheld and web platforms to reduce response times are required to enhance secu-
rity. It emphasizes perceiving the user’s perceived load and working memory load efciency. Visu-
alizations are optimized when the perceived load is reduced, and working memory increases [7]. In
Cloud computing, security-as-a-service is the most signicant change in the eld of information
and corporate security. The accuracy of live streaming data depends on several factors.
Consequently, many high-technology security systems are available. The Security Information
and Event Management (SIEM) system was developed to collect, analyze, aggregate, normalize,
stock, and purify event logs. They also correlate data from traditional security systems such as
intrusion detection/avoidance, rewalls, anti-malware, and others installed in both the host and
network domains [8].
2 Previous Studies
Almotaeryi conducted research on automated CCTV surveillance [9] in which he compared
different solutions to data augmentation in image classication using deep learning. Perez et al.
developed a method in which a neural network could learn augmentations to improve classication
process, called neural augmentation. There is an extensive discussion of this technique’s pros
and cons in different datasets [10]. Deperlıo˘
glu analyzed an effective and successful method to
diagnose diabetic retinopathy from retinal fundus images using image processing and deep learning
techniques. CNN is used to classify images [11]. In the information communication system, the
denoising of the image is a modern and practical approach due to the image-ltering algorithm.
However, that algorithm is not always efcient for the nature of the noise spectrum. Sheremet
et al. presented the possibilities of denoising using the convolutional neural network while trans-
ferring the graphical content in the information communication system proposed in this study.
They concluded that using a denoising convolutional neural network creates the correct signal but
sends noisy images [12].
Zhang et al. [13] presented the feedforward De-noising Convolutional Neural Network
(DnCNN) as the primary source of progress in deep architecture, learning algorithms, and
regularization techniques. Training processes accelerate because of residual learning and batch
normalization strategies. Zhang et al. have presented a DnCNN model with a residual learning
strategy that accommodates many images of denoising processes. Cha et al. [14] discussed a vision-
based technique for recognizing concrete cracks without calculating imperfect features with the
help of CNN architecture. However, CNN can learn image features automatically. This technique
works for removing features without the help of Image Processing Techniques (IPT) and analyzes
the efciency of CNN using Traditional Sobel and Canny edge recognition techniques, which are
signicantly better.
Yan g et al . [15] presented a new technique for super-resolution called multi super-resolution
convolutional neural network. The development of GoogLeNet architecture inspired it. This
method uses parallel convolution lters of various sizes and achieves low-resolution license plate
imagery, with a concatenation layer that blends the features. Finally, this method rebuilt the
high-resolution image using nonlinear mapping.
2588 CMC, 2021, vol.67, no.2
Tsoutsa et al. [16] are working on artistic styles using a neural algorithm that can separate and
recombine the image material and natural image style. Recently, Leon A. Gates and Alexander S.
Ecke described Image Style Transfer Using Convolutional Neural Networks and feature extraction.
Handa et al. [17], Simonyan et al. [18] worked on convolutional network depth and its impact
on the precision of large-scale image settings. They used tiny (3 ×3) convolutional lters, which
improved earlier art congurations by pushing depth to 16–19 weight layers.
Deep learning techniques play a vital role as an essential alternative, overcoming the dif-
culties of feature-based approaches. Araújo et al. [19] claried a method that classied the
hematoxylin and eosin-stained breast biopsy images using CNN. A faster R-CNN model has
been trained on the bigger datasets [20]. In the last two years, deep learning methods have been
improved rapidly for general object detection. Various methods of facial recognition are still
based on R-CNN, resulting in limited accuracy and processing speed. Jiang et al. [21]analyzed
the Faster CNN implementation, which produced impressive results in various object detection
benchmarks [17,22].
Automated Intelligence Approaches have recently been used in various scientic elds such as
medical domains, Smart cities, and health [23–27]. CNN has had a signicant impact in various
elds such as medical imaging, agriculture, smart home, smart transportation, security, and trafc
law violation detection [28–32].
3 Proposed Intelligent Ammunition Detection and Classication (IADC) System
The proposed Intelligent Ammunition Detection and Classication (IADC) system uses a
Convolutional Neural Network. Fig. 1 illustrates the acquired image stream from various CCTV
cameras at different locations. These CCTV cameras transmit captured images to the object layer
through the Cloud. Due to moving objects, the captured images may be blurred or noisy. To
convert the captured images into high-quality images, a preprocessing layer is required for image
enhancement. The object layer further sends the images to the preprocessing layer. The prepro-
cessing layer sends these image streams to the Convolution Neural Network model. The CNN
model classies the object as either with ammunition or without ammunition. If ammunition is
detected, the model noties the observer of the object and sounds an alarm. If no weapon or
ammunition is detected, there is no alarm.
3.1 Sensing Layer
There are N-sensing layers located in different positions, and each layer contains multiple
cameras. The N-sensing layers transmit captured image streams to the object layer through
the Cloud.
3.2 Object Layer
The captured image stream needs to be stored in a specic location. Image streams coming
from different sensing layers are stored in object layers. The object layers combine all stream data
inputs at a single point.
3.3 Pre-Processing Layer
Input image stream may contain noise and blurriness as a result of low quality. It consists of
raw shape data that cannot produce good results in image classication. The preprocessing layer
transforms the raw images into high-quality images by removing noise and blurriness. There are
CMC, 2021, vol.67, no.2 2589
different lters used to remove this noise and blurriness, which are the inputs of CNN. Fig. 2
shows the preprocessing process of the input raw data stream.
Figure 1: Intelligent ammunition detection & classication (IADC) system using a convolutional
neural network
Figure 2: Pre-processing image data
3.3.1 Image Noise Model
Image streaming may be blurred or noisy. The additive and multiplicative Noisy Image Model
has been claried in Eqs. (1),(2).
Ψ(x)=Φ(x)+Ψ(x)(1)
Ψ(x)=Φ(x)∗Ψ(x)(2)
where, (x) is the original image form, (x) is the noise, and (x) is the noisy image.
2590 CMC, 2021, vol.67, no.2
3.3.2 Gaussian Noise
The Gaussian noise model is very popular because of its simple application. When other noise
models fail, the Gaussian noise model can be applied. Eq. (3) is the mathematical representation
of the Gaussian noise model.
Φ(x)=1
√2πσ2e−(x−μ)2
2σ2(3)
where x is the gray value, σis the standard deviation, and μis the mean.
3.3.3 Impulse Valued Noise
The black and white dots on the image are called salt and pepper or impulse valued noise.
In Fig. 3, the centered value 200 is replaced by the value 0. Progressively, dark pixel values are
replaced by white pixel values and vice versa.
Figure 3: Pixel value is corrupted by salt and pepper noise
3.4 Convolutional Neural Network Model
Today, authorities attempt to resolve most issues by seeking help from computer professionals
using Articial Intelligence (AI) methods. AI is a broad spectrum used in every aspect of life.
For this purpose, machine learning, a subset of AI, is used. Machine learning is used to solve
a different problem by applying various algorithms like k-nearest neighbors, linear regression,
decision trees, logistic regression, Support Vector Machine, random forests, and neural networks.
Deep learning is using for image classication, a subset of machine learning. In deep learning,
CNN has been used. It is a powerful model for object classication. It is a network of different
sequentially-connected layers.
These are convolution layers in which the convolution process occurs. It can typically have
multiple convolution and pooling. Normalization layers do not necessarily follow the order. Fig. 4
shows the complete model used in the IADC system with a CNN.
To extract the feature or object for the next layer in the convolutional layer, a kernel/lter
matrix is used. There are different methods to obtain the features of the images using kernel. The
feature map values can be computed by the sum of the product of element-wise of input matrix
and kernel. Often, a dot product is used instead of the element-wise multiplication, but this can
be modied for better (or worse) results.
Fig. 5 shows the deep view of a convolution, the primary component of CNN architec-
ture that plays a vital role in feature selection from the image. Eq. (4) wasusedtocalculate
the convolution.
net(i,j)=(ξ∗ϕ)[i,j]=
m
n
ξ[m,n]ϕ[i−m,j−n](4)
CMC, 2021, vol.67, no.2 2591
Figure 4: CNN model for the proposed intelligent detection & classication (IADC) system
where net(i,j)represents the output image, ξis the input image, ϕis the kernel or lter matrix,
and * is the convolution.
Figure 5: Convolutional layer in the CNN model
The core building block of CNN is a Convolutional layer, which has been used for
feature detection.
Let’s assume the image size is a 7 ×7 matrix with RBG channels. A kernel or feature detector
or window of size 5 ×5 with three (R, G, B) channels and stride 1 is being used to scan the
kernel over the image.
In Fig. 6, if the kernel 3 ×3 moves over the image 7 ×7 matrix having stride one. Then, the
dimension of the output feature map can be calculated by Eq. (5).
Out =(W−F+2P)
S+1(5)
where, Wis the image size, Fis the kernel size, Pis padding, and Sis stride.
out =(7−3+2×0)
1+1=>5
2592 CMC, 2021, vol.67, no.2
Image
Output
Kernel / Feature Detector
Figure 6: Producing feature mapping results with 3 ×3kernelsize
Therefore, the dimension of the output feature map is 5×5. The Rectied Linear Unit (ReLU)
is an activation function, which commonly uses CNN. The ReLU function’s problem is that it is
not differentiable at the origin; therefore, it is difcult to use with backpropagation training.
It is dened mathematically, as shown in Eqs. (6),(7).
y=max(0, x)(6)
d
dxReLU(x)={One if x >0: 0 otherwise}(7)
Fig. 7 shows the graphic representation of ReLU, which creates a liner for all positive values
of x,andy=0 for all negative values of x. This activation function always produces positive and
zero outputs; if the output is negative, it converts to zero.
01234-4 -3 -2 -1
1
2
3
4
=
=0
Figure 7: ReLU of the proposed IADC system using a convolutional neural network
After the convolutional layers, the max-pooling layer was used to reduce the input stream’s
spatial dimension. The height and weight of the images were reduced. As shown in Fig. 8, the
CNN layer of the IADC system used max-pooling with 2 ×2 lter size and stride 2.
Finally, a fully connected layer becomes the input for the SoftMax layer and produces the
classication layer results.
CMC, 2021, vol.67, no.2 2593
Figure 8: Max-pooling processing with lters 2 ×2 and stride 2
3.5 Mathematical Model of CNN Loss
Y=
⎡
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎣
Y1
Y2
Y3
.
.
.
Yi
⎤
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎦
,y=
⎡
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎣
y1
y2
y3
.
.
.
yi
⎤
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎦
where the yand Yvectors represent the estimated values from the convolutional layer and initial
results. The difference between yiand Yiis called loss, which may be calculated by different
methods. The loss is used in backpropagation to update the weights. Because the calculated values
of yiare required near the original output Yi.
In the mathematical model, the aim is to backpropagate by taking the derivative of Eq. (8)
related to weights or lter ∂L
∂Wand ∂L
∂bbias. It is the cross-entropy loss that is used for classi-
cation. Because the new weights and bias will be obtained using a decent gradient algorithm by
Wnew =Wold−∝ ∂L
∂Wand bnew =b−∝∂L
∂b,∝is the learning rate parameter.
L=−
c
i=1
Yilog(yi)(8)
where c =the number of classes depending upon the implementation.
we have SoftMax conversion as in Eq. (9)
yi=eZi
n
k=1eZk(9)
2594 CMC, 2021, vol.67, no.2
where Zirepresents logits or output units, and logits will convert into probabilities via the
SoftMax conversion.
Zl=
nout
j=1
(Wjl ∗Xj)(10)
Ziis attained by interrelated weights with the Xj. Here, we get loss related to weights
contingent on dual summations in Eq. (11). One from j=1tonout and the other one from l=1
to c. Finally, the product of two derivatives will be taken
∂L
∂Wjl =
nout
j=1
c
l=1∂L
∂Zl×∂Zl
∂Wjl (11)
where ∂yi
∂Zl
is the SoftMax derivative.
In Eq. (8), loss considering yias its parameter is obliquely associated to Ziin connection
with subsequent Eq. (12).
yi=eZi
c
k=1eZk(12)
Zl=nout
j=1(Wjl ∗Xj)is given as Zi=Zl
Two demonstrations are vital, where demonstration 1, i=l, and demonstration 2 i= lwhen
i=lth unit. lis the unique neuron and pivot in SoftMax output neurons; and lneuron has the
highest values, and the rest are close to zero.
Case 1 (i=l): Now derivative of Eq. (9) via the quotient rules with reference to Zl
∂yi
∂Z(i=l)=eZic
k=1eZk−eZieZl
c
k=1eZk×c
k=1eZk
Taking common eZi
c
k=1eZk, we get the following:
∂yi
∂Zl=eZi
c
k=1eZkc
k=1eZk−eZl
c
k=1eZk
By dividing, we get
∂yi
∂Zl=eZi
c
k=1eZk1−eZi
c
k=1eZk{∵i=l}
Because we know that yi=eZi
c
k=1eZk, the above equation can be written as
∂yi
∂Zl=yi(1−yi)=yl(1−yl)for (i=l)(13)
CMC, 2021, vol.67, no.2 2595
When i= lth unit, then it has a low probability but when lis the single neuron pivot in
SoftMax output neuron. Therefore,
Case 2 (i=l): Now derivative of Eq. (12) via quotient rules with reference to Zl
∂yi
∂Zl=
∂
∂ZleZi∗c
k=1eZk−eZl∂
∂Zlc
k=1eZk
c
k=1eZk∗c
k=1eZk
It can be written as
∂yi
∂Zl=0−eZi∗eZl
c
k=1eZk∗c
k=1eZk=− eZi
c
k=1eZk∗eZi
c
k=1eZk
Because we know that yi=eZi
c
k=1eZkand yl=eZl
c
k=1eZk, we can drive the equation as
∂yi
∂Zl=−yiylfor (i= l)(14)
We can summarize Eqs. (13),(14).
∂yi
∂Zl=yi(1−yi)
−yiyl(15)
Because cross-entropy has no component of Zl, the partial derivative of Zlrelated to log(yk)
will be observed
L=−
c
i=1
(Yi∗log(yi))
Taking the derivative cross-entropy loss, the equation becomes
∂L
∂Zl=−
c
i=1YK∗∂
∂Zl
log(yk)
∂L
∂Zl=−
c
i=1
YK∂
∂yk
log(yk)∂yk
∂Zl
∂L
∂Zl=−
c
i=1
YK
yk
∂yk
∂Zl
(16)
∂yk
∂Zl
was earlier computed for the SoftMax gradient. Two demonstrations are there i= l,
k= las in Eq. (15) and Eq. (16) will be divided into two parts ∂L
∂Zl=−
YK
yk∗yk(1−yl)−
c
k=l−YK
yk∗ykylWe can simplify this as
2596 CMC, 2021, vol.67, no.2
∂L
∂Zl=−YK(1−yl)+
c
k=l
YKyl
We can further simplify this as
∂L
∂Zl=−YK+YKyl+
c
k=l
YKyl
∂L
∂Zl=yl⎛
⎝YK+
c
k=l
YK⎞
⎠−YK
where YK+c
k=lYKrepresents 1
∂L
∂Zl=yl−YK
∂L
∂Zl=yl−Yl{∵k=l}
Now, taking the derivative of Eq. (10),∂Zl
∂Wjl
we get
∂Zl
∂Wjl =xj
Now, putting the values of ∂L
∂Zl
and ∂Zl
∂Wjl
in Eq. (11),
∂L
∂Wjl =
nout
j=1
c
l=1∂L
∂Zl×∂Zl
∂Wjl
Therefore,
∂L
∂Wjl =
nout
j=1
c
l=1
(yl−Yl)xj(17)
Eq. (17) is the derivative of loss related to weights for the fully connected layer. Once ∂L
∂Wjl
is obtained by applying gradient descent on the fully connected layer; the updated weights will
be achieved.
CMC, 2021, vol.67, no.2 2597
4 Simulation and Results
MATLAB was used to simulate the proposed IADC system using a CNN. The dataset used
for the simulation contained 920 images showing persons with and without weapons. The dataset
was further divided into training (700) and validation (220). Figs. 9 and 10 show the accuracy and
extent of loss in the proposed system. Training accuracy achieved a level of 99.41%; validation
accuracy was 96.74%.
Figure 9: Performance analysis of the proposed IADC System using CNN related to iterations
and accuracy
Figure 10: Performance analysis of the proposed IADC system using CNN related to iteration
vs. loss
Fig. 10 shows the training and validation performance of the proposed IADC system related
to iteration and loss. We observed that the proposed system resulted in loss rates of 0.01 during
training and 0.09 during validation.
Fig. 11 shows the randomly selected labeled output images based on the proposed system.
The results showed that the proposed IDAC system classied persons into two classes: With
ammunition and without ammunition.
2598 CMC, 2021, vol.67, no.2
Figure 11: Randomly selected sample images from the dataset
Tab. 1 compares the performance of the proposed IADC system with previously published
models. The results show that, of the previously published models, the Overfelt-3 system [5]
offered the highest precision with 93% during training and 89% during testing. In comparison, the
proposed IADC system operated at 99.41% during training and 96.74 % during testing. Moreover,
the IADC system provided more accurate results than previously published methods like VGGNet,
OverFeat-1, OverFeat-2, and OverFeat-3 [5].
Table 1: Comparison of the IADC system using CNN with previously published models
Model Train Precision (%) Test Precision (%)
VGGNet [5]5746
OverFeat-1 [5]6256
OverFeat-2 [5]6964
OverFeat-3 [5]9389
Proposed IADC system using CNN 99.41 96.74
CMC, 2021, vol.67, no.2 2599
5 Conclusions and Future Studies
In conclusion, the Convolutional Neural Network application in security systems offers more
precise detection of armed persons and weapons. Additionally, it has given more accurate results
than previously published methods such as VGGNet, Overfeat-1, Overfeat-2, and Overfeat-3. The
proposed IADC system achieved a 96.74% accuracy rate and a 3.26% loss rate.
In future studies, the Yolo model can be used to obtain more precise results through compar-
ing results. Moreover, improved precision in detecting and classifying a wider variety of weapons
and ammunition can be achieved.
Acknowledgement: Thanks to our families and colleagues, who provided moral support.
Funding Statement: The authors received no specic funding for this study.
Conicts of Interest: The authors declare that they have no conicts of interest to report regarding
this study.
References
[1] M. Grega, A. Matiola´
nski, P. Guzik and M. Leszczuk, “Automated detection of rearms and knives
in a CCTV image,” Sensors, vol. 16, no. 47, pp. 1–16, 2016.
[2] R. Olmos, S. Tabik and F. Herrera, “Automatic handgun detection alarm in videos using deep
learning,” Neurocomputing, vol. 275, pp. 66–72, 2018.
[3] A. Atta, M. A. Khan, S. Abbas, G. Ahmad and A. Fatimat, “Modelling smart road trafc con-
gestion control system using machine learning techniques,” Neural Network World, vol. 29, no. 2,
pp. 99–110, 2019.
[4] Ahmed, “Detection of objects in images using region-based convolutional neural networks,” United
States Patent, vol. 2, pp. 1–20, 2018.
[5] J. Lai and S. Maples, “Developing a real-time gun detection classier,” Technical Report, vol. 2017,
pp. 1–4, 2017.
[6] A. Caragliu and C. F. D. Bo, “Smart innovative cities: The impact of smart city policies on urban inno-
vation,” Technological Forecasting and Social Change, vol. 142, no. December 2017, pp. 373–383, 2019.
[7] N. Cowan, “The many faces of working memory and short-term storage,” Psychonomic Bulletin &
Review, vol. 24, no. 4, pp. 1158–1170, 2017.
[8] A. Ahi and A. V. Singh, “Role of distributed ledger technology (DLT) to enhance resiliency in internet
of things (IOT) ecosystem,” in Amity Int. Conf. on Articial Intelligence, Dubai, United Arab Emirates,
pp. 782–786, 2019.
[9] Almotaeryi and Resheed, “An integrated framework for CCTV infrastructures deployment in KSA:
Towards an automated surveillance,” Doctoral Thesis, North Umbria University, pp. 1–183, 2018.
[10] L. Perez and J. Wang, “The effectiveness of data augmentation in image classication using deep
learning,” Arxiv Preprint Arxiv, vol. 2017, pp. 1–8, 2017.
[11] M. I. Razzak, S. Naz and A. Zaib, “Deep learning for medical image processing: Overview, challenges
and the future,” Lecture Notes in Computational Vision and Biomechanics, vol. 26, pp. 323–350, 2018.
[12] O. Sheremet, K. Sheremet, O. Sadovoi and Y. Sokhina, “Convolutional neural networks for image
denoising in infocommunication systems,” in Int. Scientic-Practical Conf. Problems of Infocommunica-
tions Science and Technology, Kharkiv, Ukraine, pp. 429–432, 2019.
[13] K. Zhang, W. Zuo, Y. Chen, D. Meng and L. Zhang, “Beyond a Gaussian denoiser: Residual learning
of deep CNN for image denoising,” IEEE Transactions on Image Processing, vol. 26, no. 7, pp. 3142–
3155, 2017.
2600 CMC, 2021, vol.67, no.2
[14] Y. J. Cha, W. Choi and O. Büyüköztürk, “Deep learning-based crack damage detection using con-
volutionalneuralnetworks,”Computer-Aided Civil and Infrastructure Engineering, vol. 32, no. 5,
pp. 361–378, 2017.
[15] Y. Yang, P. Bi and Y. Liu, “License plate image super-resolution based on convolutional neural
network,” in IEEE 3rd Int. Conf. on Image, Vision and Computing, Chongqing, China, pp. 723–727, 2018.
[16] P. Tsoutsa, P. Fitsilis, L. Anthopoulos and O. Ragos, “Nexus services in smart city ecosystems,” Journal
of the Knowledge Economy, vol. 11, no. 3, pp. 1–20, 2020.
[17] A. Handa, P. Garg and V. Khare, “Masked neural style transfer using convolutional neural networks,”
in Int. Conf. on Recent Innovations in Electrical, Electronics & Communication Engineering, Bhubaneswar,
India, pp. 2099–2104, 2018.
[18] K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,”
ArXiv Preprint ArXiv, vol. 2014, pp. 1–14, 2014.
[19] V. Araujo, K. Mitra, S. Saguna and C. Åhlund, “Performance evaluation of ware: A cloud-based IoT
platform for smart cities,” Journal of Parallel and Distributed Computing, vol. 132, pp. 250–261, 2019.
[20] R. Girshick, J. Donahue, T. Darrell, J. Malik, U. C. Berkeley et al., “Deformable part models are
convolutional neural networks,” in IEEE Conf. on Computer Vision and Pattern Recognition, Boston,
USA, pp. 437–446, 2015.
[21] H. Jiang and E. L. Miller, “Face detection with the faster R-CNN,” in 12th IEEE Int. Conf. on
Automatic Face & Gesture Recognition, Washington, USA, pp. 650–657, 2017.
[22] Y. Hua, R. Li, Z. Zhao, X. Chen and H. Zhang, “Gan-powered deep distributional reinforcement learn-
ing for resource management in network slicing,” IEEE Journal on Selected Areas in Communications,
vol. 38, no. 2, pp. 334–349, 2020.
[23] A. Atta, S. Abbas, M. A. Khan, G. Ahmed and U. Farooq, “An adaptive approach: Smart trafc
congestion control system,” Journal of King Saud University-Computer and Information Sciences,vol.32,
no. 9, pp. 1012–1019, 2018.
[24] M. Asadullah, M. A. Khan, S. Abbas, A. Athar, S. S. Raza et al., “Blind channel and data esti-
mation using fuzzy logic-empowered opposite learning-based mutant particle swarm optimization,”
Computational Intelligence and Neuroscience, vol. 2018, no. 5, pp. 1–12, 2018.
[25] K. Iqbal, M. A. Khan, S. Abbas, Z. Hasan and A. Fatima, “Intelligent transportation system (ITS) for
smart-cities using mamdani fuzzy inference system,” International Journal of Advanced Computer Science
and Applications, vol. 9, no. 2, pp. 94–105, 2018.
[26] G. Ahmad, M. A. Khan, S. Abbas, A. Athar, B. S. Khan et al., “Automated diagnosis of hepatitis B
using multilayer mamdani fuzzy inference system,” Journal of healthcare engineering, vol. 2019, no. 1,
pp. 1–11, 2019.
[27] S. Zahra, M. A. Khan, M. N. Ali and S. Abbas, “Standardization of cloud security using mamdani
fuzzier,” International Journal of Advanced Computer Science and Applications, vol. 9, no. 3, pp. 292–
297, 2018.
[28] Gu and Jingtian, “The application of convolutional neural network in security code recognition,”
Journal of Physics: Conference Series, vol. 1187, no. 4, pp. 42064, 2019.
[29] Q. Chen, X. Gan, W. Huang, J. Feng and H. Shim, “Road damage detection and classication using
mask R-CNN with densenet backbone,” Computers, Materials & Continua, vol. 65, no. 3, pp. 2201–
2215, 2020.
[30] R. Chen, L. Pan, C. Li, Y. Zhou, A. Chen et al., “An improved deep fusion CNN for image
recognition,” Computers Materials & Continua, vol. 65, no. 2, pp. 1691–1706, 2020.
[31] Y. Chen, X. Qin, L. Zhang and B. Yi, “A novel method of heart failure prediction based on
dpcnnxgboost model,” Computers Materials & Continua, vol. 65, no. 1, pp. 495–510, 2020.
[32] Y. Zhao, J. Cheng, P. Zhang and X. Peng, “ECG classication using deep CNN improved by wavelet
transform,” Computers Materials & Continua, vol. 64, no. 3, pp. 1615–1628, 2020.
