IAES International Journal of Artificial Intelligence (IJ-AI)

Vol. 11, No. 2, June 2022, pp. 405~413

ISSN: 2252-8938, DOI: 10.11591/ijai.v11.i2.pp405-413  405

Vehicles detection and counting based on internet of things

technology and video processing techniques

Marwa A. Marzouk1, Amr Abd El Azeem2

1
Department of Information Technology, Matrouh University, Mersa Matruh, Egypt
2
Milatery Technical College, Computer Engineering, Cairo, Egypt

Article Info ABSTRACT

Article history: Recent studies have proven that vehicle tracking and detection play an
important role in traffic density monitoring. Traffic overcrowding can be
Received Sept 9, 2021 effectively controlled if the number of vehicles expected to pass through a
Revised Feb 9, 2022 congested intersection can be predicted ahead of time. To overcome such
Accepted Feb 21, 2022 impact of traffic congestion the proposed system presents a framework,
using motion detection algorithms and “ThingSpeak” internet of things (IoT)
platform which is used in to calculate traffic density, the proposed system
Keywords: capturing video with wireless internet protocol (IP) cameras and
broadcasting it to the server where motion detection algorithms as
Image processing background subtraction are used to obtain a quick overview of traffic
Internet of things density, To save cost and improve the solution, the suggested system utilizes
Motion detection image processing techniques as well as the IoT analytic platform
Traffic density “ThingSpeak” to monitor traffic density. Finally, the suggested method is
used to manage traffic flow and avoid traffic crowded. The results of the
studies show that the integration of IoT-based technologies with a modified
background subtraction technique is effective. This method might be
enhanced further to detect vehicles that break traffic laws. We may also
improve this system by detecting the presence of emergency vehicles
(including an ambulance or fire truck) and granting priority to those cars.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Marwa A. Marzouk
Department of Information Technology, Matrouh University
Mersa Matruh, Egypt
Email: [email protected]

1. INTRODUCTION
As a result of the growing population, there is a greater demand for vehicles, which increases traffic
congestion in various places [1], [2]. So reliable vehicle detection is a critical step of vehicle recognition a
smart traffic system can improve, road safety and reduce traffic congestions. For on-road vehicle detection,
imaging technology has advanced significantly. Nowadays, cameras are available at a lower cost., are
compacted and of high quality. Vehicle detection advances using the internet of things (IoT) have proven
their utility in practically every aspect of our everyday lives and can be regarded as a tool for traffic
management via a central server [3], [4].
Because of, image-databases and live video information are becoming more and more, because of,
image-databases and live video information are becoming more and more, Widespread and intelligent, they
are exceptionally important [5], [6]. For the monitoring of traffic data. image and video processing modules
are utilized, with motion detection algorithms that identify automobiles as moving blobs and track them for
several frames difference, and a modified background subtraction technique is employed in this paper for
daylight vehicle recognition. Through frame pre-processing using smoothing with red green blue )RGB( data

Journal homepage: http://ijai.iaescore.com

406  ISSN: 2252-8938

format to remove noise and unwanted objects such as rain and clouds and to get more information from the
frame. Also, the pre-processed frames are used for background modeling to provide a statistical description
of the entire background scene using non-recursive frame differencing technique with 4 frames difference
and scaled frame to minimize the modeling time. To extract foreground objects using 2 pre-calculated
thresholds (strong threshold and weak threshold) not only one threshold to give an accurate decision about
moving vehicles in the video sequence then count the number of vehicles.
Finally, the proposed system intends to construct an automatic vehicle counting system that can
analyze video captured using wireless internet protocol (IP) cameras on roadways near traffic
intersections/junctions, followed by vehicle counting. Data from traffic monitoring will be analysed using the
ThingSpeak channel. ThingSpeak channel is an IoT analytics platform with the ability to collect real-time
data and visualize the data in the form of charts. We build an IoT-based traffic signal monitoring and control
system image processing on this framework. For traffic density monitoring, manual signaling mode, and
automatic signaling mode, the system is developed and simulated. IoT technology has recently been applied
in a wide range of applications.
In this section, we'll look at some of the work that's been done using IoT in traffic management. A
great deal of research is being done to address the issues of vehicle detection and tracking [7]. In [8] to
establish the expected needed timing of traffic lights, an array of infrared sensors is used to count the number
of vehicles on each traffic lane of the road and record the information on the cloud via a bluetooth
connectionuses clustering techniques based on the k-nearest neighbors (KNN) algorithm. Bluetooth needs
data transfer from access points close to the sensor array, which adds to the system’s complexity. Using the
clustering technique also adds to the cloud computing system’s overhead, causing delays in decision-making
and traffic light timing changes. night vehicle detection approach represented by Nath and Deb [3] as
template matching because it contains information about the number of libraries of templates, the approach is
ineffective, and approximating correlation is a challenging process. Kim et al. [9] and Zhang et al. [10]
proposed approaches for night vehicle identification based on tracking and lighting pairing. an IoT-based
traffic control system was proposed in [11] for image processing, this system utilizes MATLAB software,
and for delivering vehicle data sets, it used a Wi-Fi module. If direct data transfer through the cloud is
utilised instead of a Wi-Fi transceiver, the system may be made more effective. In [12] a traffic control
system was built that utilized a wireless transmitter to send photos directly to the main server. Then there’s
the server, where the process is more effective if the data being communicated isn’t in the form of images.
Rather, the processed output information is delivered directly, saving a significant amount of time and
communication weight. In [13] the use of networking and embedded technologies to solve traffic congestion
problems is described. Using Raspberry Pi, routers, an ultrasonic sensor, and email servers, the author created
an alarm system. The intelligent traffic system model was presented by Badura and Lieskovsky [14] the
cameras positioned at the intersections scan and monitor their respective domains. For general image
analysis, the collected data is instantly transferred to a topology independent data delivery system
photoelectric sensors, according to Salama et al. [15] might be used to regulate traffic lights. The precise
locations for sensor placement are one of the most important aspects. This is mostly due to the traffic control
department’s need to track cars’ movements at specified times, particularly during busy times.

2. RESEARCH METHOD
A new method is being developed to reduce the effects of traffic congestion by combining image
processing and IoT technology. This system uses wireless IP cameras at traffic intersections to capture video
and send it to a server, where the vehicle density on the road is computed using video and image processing
techniques. The total number of cars will be spotted and estimated after obtaining a quick look at traffic
conditions. The suggested system was developed by using the MATLAB programming environment. The
ThinSpeak channel would be used to do traffic control analyses [16]. The next two subsections go through all
of the specifics of how the work was completed.

2.1. Vehicle detection using image processing

2.1.1. Modified background subtraction algorithm for vehicle detection
Detecting moving objects in a video stream is an important and important topic of computer vision
research. Background subtraction is a common method for identifying moving vehicles in a video frame
segment that clearly distinguishes from the background model. Developing a decent background subtraction
method is difficult for a variety of reasons. It should, first, be sensitive to variations in lighting. Second, it
must be avoided to detect non-stationary background noise objects for instance snow, moving leaves, rain,
and shadows cast by moving objects [17], [18]. Figure 1 represents a block diagram of video processing.

Int J Artif Intell, Vol. 11, No. 2, June 2022: 405-413

Int J Artif Intell ISSN: 2252-8938  407

Figure 1. Flow diagram of video processing

Function Main_Function()
For each Frame in FarmesSequence
Call FramePreprocessing(Frame)
Call FrameScaling(Frame)
Frame_Number:=Frame_Number +1
Add Frame to Frame_Array_List
If Frame_Number>4 then
Loop
I:=0
Get Last 4 Frames
New_Frame=Average(Last_4_Frames)
Add New_Frame to New_Frame_Array_List
If New_Frame_Array_List.Count()>2 then
Call Frame_Differencing(New_Frame_Array_List[i]
New_Frame_Array_List[i+1])
I:=1+1
End if
End Loop
End If
End Loop
End Main Function

a. Preprocessing
To increase the detection of moving vehicles, the video must be pre-processed. To get more
information from the frame by using smoothing with RGB data format to remove noise, Figure 2 in this
paper for instance, by spatial and temporal smoothing, snow can be cut out of a video as shown in
Figure 2(a). After the recognition of small moving objects, morphological processing of the frames can be
used to remove them as shown in Figure 2(b). Such as moving leaves on a tree as shown in Figure 3 by
comparing video frame that shows the objects detected as moving in Figure 3(a) with video frame in
Figure 3(b) after excluding the moving objects by using the morphological processing.

(a) (b)

Figure 2. Comparing the video frame in (a) while it snowed in the traffic scene with the same video frame in
(b) after excluding the snow streaks after spatial and temporal smoothing

Vehicles detection and counting based on internet of things technology and video … (Marwa A. Marzouk)
408  ISSN: 2252-8938

(a) (b)

Figure 3. Comparing video frame that shows the objects detected as moving in (a) with video frame in
(b) after excluding the moving objects by using the morphological processing

Function FramePreprocessing(InputFrame) returnes Frame

Set OutputFrameArray[frame_width][frame_height] edgex=(frame_width/2)
edgey==(frame_height/2)
for x from edgex to frame_width-edgex
for y from edgey to frame_height-edgey
Set colorArray[frame_width][frame_height]
for fx from 0 to frame_width
for fy from 0 to frame_height
colorArray[fx][fy]:=Inputframe[x+fx-edgex][y+fy-edgey]
sort all entries in colorArray[][]in ascending prder
OutputFrameArray [x][y]:=colorArray[farme_width/2][frame_height/2]
Return Frame
End FramePreprocessing Function

b. Background modeling
Background modeling uses the new video frame to calculate and update a background model.
Recursive techniques do not maintain a buffer for background estimation. Compared with non-recursive
techniques, recursive techniques require less storage, but any error in the background model can linger for a
much longer period of time. The preprocessed frames are used for background modeling to provide a
statistical description of the entire background scene, using non-recursive frame differencing technique with
4 frames difference and scaled frame to minimize the modeling time to extract foreground objects and, using
2 pre-calculated thresholds (strong threshold and weak threshold) not only one threshold to give an accurate
decision about moving objects in the video sequence.

c. Foreground detection
The background model is compared to the input video frame, and relevant foreground pixels out
from the input frame are identified. Then, foreground detection finds pixels in the video frame that the
background model can’t explain and outputs them as a binary relevant foreground mask. The most standard
method for detecting foreground objects is to see if the input pixel differs considerably from the equivalent
background estimate:

|𝐼𝑡 (x; y)– 𝐵𝑡 (x; y) | > T (1)

A threshold based on normalized data is another common foreground detection method:

| 𝐼𝑡 (x,y) – 𝐵𝑡 (x,y) − 𝜇𝑑 |
> Ts (2)
∂𝑑

Where µd and ∂d are the mean and the standard deviation of It (x, y)-Bt (x, y) for all spatial locations (x, y).
Most schemes determine the foreground threshold T or Ts experimentally. Ideally, the threshold should be a
function of the spatial location (x, y). For instance, in low-contrast areas, the threshold should be lower. To
accentuate the contrast in dark areas such as shadows, one possible change is to use relative difference
instead of absolute difference:

| 𝐼𝑡 (x,y) – 𝐵𝑡 (x,y) |
𝐵𝑡 (x,y)
> Tc (3)

Int J Artif Intell, Vol. 11, No. 2, June 2022: 405-413

Int J Artif Intell ISSN: 2252-8938  409

Using two thresholds is another way to introduce spatial variability. The aim is to detect “strong”
foreground pixels with absolute deviations from background estimations that exceed a specific threshold.
Then, from strong foreground pixels, foreground regions are produced by integrating surrounding pixels with
absolute differences greater than a lower threshold. To expand the area, a two-pass, connected-component
grouping method might be applied.

d. Data validation
In this step, the candidate mask is examined, and pixels that do not relate to actual moving objects
are removed before the final foreground mask is produced. As shown in Figure 4 where the highway
sequence processed with a background subtraction method and our method Figure 4(a) input frame,
Figure 4(b) background subtraction, and Figure 4(c) motion objects. While most of the background models
have three main problems [19], [20]:
i) They ignore any pixel-to-pixel connection. Small false-positive or false-negative portions are irregularly
spread across the candidate mask as a result of this problem. To eliminate these portions, common
methods use morphological filtering and related component grouping [21].
ii) The rate of background modeling adaptation may not match the moving speed of the foreground
objects. To solve these issues by running numerous background models at varying adaption rates and
cross-validating between these models regularly to improve accuracy.
iii) Moving leaves or moving object shadows cast non-stationary pixels that are simply mistaken for actual
foreground objects. By using sophisticated background modeling techniques like mixture of Gaussians
(MoG) and morphological filtering for cleanup, the moving leaves problem can be solvedas shown in
Figure 4(b).
iv) Finally, only doing background subtraction on a sub-sampled version of each image is a simplification
strategy that speeds up data validation. As a result, a 640 pixel by 480-pixel image can be resized to 160
by 120 pixels, a fraction of the linear dimensions. Because this image is one-sixteenth the size of the
original, the background subtraction processing time is also one-sixteenth the time as shown in
Figure 4(c).

(a)

(b) (c)

Figure 4. The highway sequence processed with a background subtraction methode and our method (a) input
frame, (b) background subtraction, and (c) motion objects
Vehicles detection and counting based on internet of things technology and video … (Marwa A. Marzouk)
410  ISSN: 2252-8938

e. Traffic count
In a binary image, blobs are the linked sections. The goal of the blob analysis method is to find spots
and/or areas in a picture that differ in brightness or area [22]. As explained in [23] the laplacian of Gaussian
is utilized as a formulation for finding blob value using a computer vision approach. The procedure begins
with the labeling of an area that is deemed a foreground object, followed by the collection of data into the
blob, such as the initial pixel location, x- and y-axis lengths, and pixel area. Figure 5 illustrates a blob area. In
Figures 5(a) to (c) are visible object detection processes that are acquired by foreground area detection from
the binarization process as Figure 5(d). Setting pixel vector values to point values can help in determining
blob area as demonstrated in Figure 5(e).

(a)

Figure 5. Vechicles counting (a) big frame of the original image, (b) cropping object, (c) foreground
segmentation, (d) object detection using the bounding box, and (e) BLOB area in x, y-axis direction

2.1.2. Traffic density monitoring on IoT analytics platform

The vehicle density count is provided by joining up for ThingSpeak and creating a channel. Figure 6
shows how the traffic signal timer is set based on the number of vehicles. We added a description and a field
to the channel settings before it could be used to contain data [24]. A write accept payments online (API) key
is generated by default when we build a channel. When transmitting data to the channel the write API key is
used. Figure 7 displays the time and date of the road congestion monitoring analysis on a ThingSpeak
channel. Each dot represents the current traffic value and the moment in which this was sent to the channel.

Figure 6. Sign-up windows for ThingSpeak

Int J Artif Intell, Vol. 11, No. 2, June 2022: 405-413

Int J Artif Intell ISSN: 2252-8938  411

Figure 7. Traffic values

3. RESULTS AND DISCUSSION

The proposed method was tested on 8 videos captured using an IP webcam. Measurements in this
paper were made in different environments to measure the strength of the algorithm under factors that will
decrease the accuracy and performance of the algorithm like moderate rainfall as shown in Table 1, fog as
shown in Table 2, and night as shown in Table 3. The data is analyzed by thingSpeak which is an IoT
analytics platform with the capability of collecting real-time data and visualizing it in the form of charts. To
calculate accuracy in this paper we use the standard accuracy equation [25]
Accuracy=(TP+TN)(TP+TN+FN+FP)x100%. Where TP is a true positive, TN is a true negative, FP is a false
positive, and FN is a false negative.

Table 1. Performance of the traffic monitoring system in moderate rainfall

Vid.NO Vehicles Count TP TN FP FN Accuracy
1 6 2 2 0 0 100
2 10 2 2 0 1 80
3 8 3 2 0 1 83
4 9 2 4 1 0 85.7
5 11 4 5 1 1 90
6 2 1 1 0 0 100
7 4 2 1 0 0 100
8 5 1 3 0 0 100
Average 93

Table 2. Performance of the traffic monitoring system in Fog

Vid.NO Vehicles Count TP TN FP FN Accuracy
1 5 1 2 0 0 100
2 4 2 4 0 1 85.7
3 8 3 2 1 0 83
4 7 4 5 1 1 90
5 10 2 2 0 1 80
6 3 1 1 0 0 100
7 9 2 3 1 0 83
8 12 1 3 1 0 80
Average 88

Table 3. Performance of the traffic monitoring system in night

Vid.NO Vehicles Count TP TN FP FN Accuracy
1 5 4 4 2 0 80
2 4 2 4 0 1 85.7
3 8 3 6 1 2 81.8
4 7 3 4 1 1 77.7
5 10 5 4 0 2 81.1
6 3 2 2 1 0 80
7 9 2 3 1 0 83
8 12 6 4 0 2 83.3
Average 82

Vehicles detection and counting based on internet of things technology and video … (Marwa A. Marzouk)
412  ISSN: 2252-8938

4. CONCLUSION
The modified background subtraction technique, noise reduction via morphological analysis, blob
detection, and a signal system based on numerous blobs or density a module were created in this study. We
also presented a solution for identifying vehicle density using an IoT analytic platform. The suggested
approach was tested on eight videos taken using an IP camera. Measurements are taken in a variety of
environments to assess the algorithm’s strength in the face of factors that might reduce accuracy and
performance such as (fog, rain, and night). And based on these measurements the accuracy of rain is about
93% also in fog is 88% however at night the accuracy is about 83% so the proposed system needs to enhance
its accuracy at night but the main advantage of the proposed system using image processing, MATLAB
software and ThingSpeak platform. that it is possible to execute it at a low cost and with the highest level of
accuracy using the suggested technique, we are just monitoring the number of vehicles present at the signal.
This method might be developed further to detect vehicles that violate traffic laws also this method might be
developed further to identify the presence of emergency vehicles (such as an ambulance or fire truck) and
give those vehicles precedence.

REFERENCES
[1] E. Bekiaris and Y. J. Nakanishi, Eds., Economic impacts of intelligent transportation systems: innovations and case studies, 1st
Editio. Elsevier, 2004.
[2] Y. A. Syahbana and Y. Yasunari, “Early detection of incoming traffic for automatic traffic light signaling during roadblock using
vanishing point-guided object detection and tracking,” in 2021 60th Annual Conference of the Society of Instrument and Control
Engineers of Japan (SICE), 2021, pp. 1466–1471.
[3] R. K. Nath and S. K. Deb, “On road vehicle/object detection and tracking using template,” Indian Journal of Computer Science
and Engineering, vol. 1, no. 2, pp. 98–107, 2010.
[4] R. P. Nimkar and C. N. Deshmukh, “Traffic density monitoring and cattle menace alert system using IoT,” International Journal
of Research in Engineering and Technology, vol. 7, no. 8, pp. 96–104, Aug. 2018, doi: 10.15623/ijret.2018.0708012.
[5] N. K. Jain, R. K. Saini, and P. Mittal, “A review on traffic monitoring system techniques,” in Advances in Intelligent Systems and
Computing, Springer Singapore, 2019, pp. 569–577.
[6] V. Kastrinaki, M. Zervakis, and K. Kalaitzakis, “A survey of video processing techniques for traffic applications,” Image and
Vision Computing, vol. 21, no. 4, pp. 359–381, Apr. 2003, doi: 10.1016/S0262-8856(03)00004-0.
[7] D. Nettikadan and S. R. M. S., “Smart community monitoring system using thingspeak iot plaform,” International Journal of
Applied Engineering Research, vol. 13, no. 17, pp. 13402–13408, 2018.
[8] S. Kumar Janahan, M. R. M. Veeramanickam, S. Arun, K. Narayanan, R. Anandan, and S. Javed Parvez, “IoT based smart traffic
signal monitoring system using vehicles counts,” International Journal of Engineering and Technology, vol. 7, no. 2.21, Art. no.
309, Apr. 2018, doi: 10.14419/ijet.v7i2.21.12388.
[9] H.-K. Kim, S. Kuk, M. Kim, and H.-Y. Jung, “An effective method of head lamp and tail lamp recognition for night time vehicle
detection,” World Academy of Science, Engineering and Technology, vol. 44, pp. 1091–1094, 2010.
[10] W. Zhang, Q. M. J. Wu, G. Wang, and X. You, “Tracking and pairing vehicle headlight in night scenes,” IEEE Transactions on
Intelligent Transportation Systems, vol. 13, no. 1, pp. 140–153, Mar. 2012, doi: 10.1109/TITS.2011.2165338.
[11] E. Basil and S. D. Sawant, “IoT based traffic light control system using Raspberry Pi,” in 2017 International Conference on
Energy, Communication, Data Analytics and Soft Computing (ICECDS), Aug. 2017, pp. 1078–1081, doi:
10.1109/ICECDS.2017.8389604.
[12] T. Osman, S. S. Psyche, J. M. Shafi Ferdous, and H. U. Zaman, “Intelligent traffic management system for cross section of roads
using computer vision,” in 2017 IEEE 7th Annual Computing and Communication Workshop and Conference (CCWC), Jan.
2017, pp. 1–7, doi: 10.1109/CCWC.2017.7868350.
[13] K. B. Malagund, S. N. Mahalank, and R. M. Banakar, “IoT based smart city traffic alert system design,” in 2016 International
Conference on Computing Communication Control and automation (ICCUBEA), Aug. 2016, pp. 1–6, doi:
10.1109/ICCUBEA.2016.7860146.
[14] S. Badura and A. Lieskovsky, “Intelligent traffic system: cooperation of MANET and image processing,” in 2010 First
International Conference on Integrated Intelligent Computing, Aug. 2010, pp. 119–123, doi: 10.1109/ICIIC.2010.41.
[15] A. S. Salama, B. K. Saleh, and M. M. Eassa, “Intelligent cross road traffic management system (ICRTMS),” in 2010 2nd
International Conference on Computer Technology and Development, Nov. 2010, pp. 27–31, doi: 10.1109/ICCTD.2010.5646059.
[16] J. M. S. Waworundeng, D. Fernando Tiwow, and L. M. Tulangi, “Air pressure detection system on motorized vehicle tires based
on IoT platform,” in 2019 1st International Conference on Cybernetics and Intelligent System (ICORIS), Aug. 2019, pp. 251–256,
doi: 10.1109/ICORIS.2019.8874904.
[17] S. González Díaz, “Fall detection system for the elderly by means of artificial vision (in Spanish).” 2017.
[18] B. Garcia-Garcia, T. Bouwmans, and A. J. Rosales Silva, “Background subtraction in real applications: Challenges, current
models and future directions,” Computer Science Review, vol. 35, Art. no. 100204, Feb. 2020, doi: 10.1016/j.cosrev.2019.100204.
[19] R. Ke, Y. Zhuang, Z. Pu, and Y. Wang, “A smart, efficient, and reliable parking surveillance system with edge artificial
intelligence on IoT devices,” IEEE Transactions on Intelligent Transportation Systems, vol. 22, no. 8, pp. 4962–4974, Aug. 2021,
doi: 10.1109/TITS.2020.2984197.
[20] A. Rosenfeld and D. Weinshall, “Extracting foreground masks towards object recognition,” Nov. 2011, doi:
10.1109/iccv.2011.6126391.
[21] Q. Zhang, J. Lin, Y. Tao, W. Li, and Y. Shi, “Salient object detection via color and texture cues,” Neurocomputing, vol. 243, pp.
35–48, Jun. 2017, doi: 10.1016/j.neucom.2017.02.064.
[22] A. Alharbi, A. Aloufi, E. Hamawi, F. Alqazlan, S. Babaeer, and F. Haron, “Counting people in a crowd using Viola-Jones
algorithm,” International Journal of Computing, Communication and Instrumentation Engineering, vol. 4, no. 1, pp. 57–59, Feb.
2017, doi: 10.15242/IJCCIE.IAE1216010.
[23] W. Liu, S. Liao, W. Ren, W. Hu, and Y. Yu, “High-level semantic feature detection: a new perspective for pedestrian detection,”

Int J Artif Intell, Vol. 11, No. 2, June 2022: 405-413

Int J Artif Intell ISSN: 2252-8938  413

in 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Jun. 2019, pp. 5182–5191, doi:
10.1109/CVPR.2019.00533.
[24] S. Pasha, “Thingspeak based sensing and monitoring system for IoT with Matlab analysis,” International Journal of New
Technology and Research, vol. 2, no. 6, pp. 19–23, 2016.
[25] A. S. More and D. P. Rana, “An experimental assessment of random forest classification performance improvisation with
sampling and stage wise success rate calculation,” Procedia Computer Science, vol. 167, pp. 1711–1721, 2020, doi:
10.1016/j.procs.2020.03.381.

BIOGRAPHIES OF AUTHORS

Marwa A. Marzouk Ph.D. Institute of Graduate Studies and Research

Department of information technology, Alexandria University, 2017. M.sc, Institute of
Graduate Studies and Research Department of information technology, Alexandria University,
2011. Post graduate diploma, Institute of Graduate Studies and Research Department of
information technology, Alexandria University, 2008. Bachelor of Science and Education-
Alexandria University.interested in Computer Vision, Image Processing, and Information
Retrieval. She can be contacted at email: [email protected].

Amr Abd El Azeem Milatery Technical College, Computer Engineering, Cairo,

Egypt. Master in Cyber security from arab academy for science, technology and maritime
transport. interested in Computer Vision, Image Processing, network security, cyber security,
internet of things and Information Retrieval. He can be contacted at email:
[email protected].

Vehicles detection and counting based on internet of things technology and video … (Marwa A. Marzouk)