See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/355886757

Face Detection and Recognition Using OpenCV

Article · October 2021

DOI: 10.30880/jscdm.2021.02.02.008

CITATIONS READS

8 10,290

2 authors:

Ramadan Zebari Amira Bibo Sallow

Duhok Polytechnic University Duhok Polytechnic University
6 PUBLICATIONS 13 CITATIONS 47 PUBLICATIONS 689 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Parallel processing View project

distributed system View project

All content following this page was uploaded by Ramadan Zebari on 29 August 2023.

The user has requested enhancement of the downloaded file.

 JOURNAL OF SOFT COMPUTING AND DATA MINING VOL.2 NO. 1 (2021) 86-97

© Universiti Tun Hussein Onn Malaysia Publisher’s Office

Journal of Soft
JSCDM Computing and
Data Mining
Journal homepage: http://penerbit.uthm.edu.my/ojs/index.php/jscdm
e-ISSN : 2716-621X

Face Detection and Recognition Using OpenCV

Ramadan TH. Hasan1*, Amira Bibo Sallow2
1
IT Department, Technical College of Informatics Akre,
Duhok Polytechnic University, Duhok, Kurdistan Region, IRAQ
2
College of Engineering,
Nawroz University, Duhok, Kurdistan Region, IRAQ

DOI: https://doi.org/10.30880/jscdm.2021.02.02.008
Received 21 May 2021; Accepted 05 August 2021; Available online 15 October 2021

Abstract: Intel's OpenCV is a free and open-access image and video processing library. It is linked to computer
vision, like feature and object recognition and machine learning. This paper presents the main OpenCV modules,
features, and OpenCV based on Python. The paper also presents common OpenCV applications and classifiers
used in these applications like image processing, face detection, face recognition, and object detection. Finally, we
discuss some literary reviews of OpenCV applications in the fields of computer vision such as face detection and
recognition, or recognition of facial expressions such as sadness, anger, happiness, or recognition of the gender and
age of a person.

Keywords: OpenCV, Face Detection, Object Detection, Eigenfaces, Faster R-CNN, Fisherfaces

1. Introduction
Computer Vision is one of the most fascinating and challenging tasks in the field of Artificial Intelligence.
Computer Vision serves as a link between computer software and the visuals we see around us. It enables computer
software to comprehend and learn about the visuals in its environment. As an example: The fruit is determined by its
color, shape, and size. This job may seem simple for the human brain, but in the Computer Vision pipeline, we first
collect data, then conduct data processing operations, and then train and educate the model to learn how to differentiate
between fruits based on size, shape, and color. The main goal is to identify and comprehend the images and offer new
images that are more useful for us in different life fields [1], [2].
The term "OpenCV" is an abbreviation for "open source computer vision." The architecture is made up of
software, databases, and plugins that are pre-programmed with support for integrating computer vision applications [3].
It is one of the most used toolkits with a large developer group. It is well-known for the size at which it builds real-
world usage cases for industrial use. OpenCV follows C/C++, Python, Java programming languages and can be used to
build computer vision software for desktop and smartphone platforms such as Windows, Linux, macOS, Android, and
iOS. The most recent releases are OpenCV-4.5.2 and OpenCV-3.4.14. It is free and open-source, as well as simple to
use and install. It is intended for numerical productivity with a heavy emphasis on real-time applications. The first
version was in the C programming language; however, its success increased with the release of Version 2.0, which had
a C++ implementation [2]. C++ is used to create new features. OpenCV can be downloaded for free from
http://opencv.org. This platform includes the most recent distribution update (version: 4.5.2) as well as older iterations.
Photos must be in BGR or Grayscale format in order to be displayed or saved via OpenCV. Otherwise, unfavorable
outcomes could occur [1].
Face detection is a form of computer vision that aids in detecting and visualizing facial features in captured
pictures or real-time videos. This type of object detection technique detects instances of semantic artifacts of a given
class (such as people, cars, and houses) in digital pictures and videos. Face recognition has become increasingly
important as technology has advanced, especially in fields such as photography, defense, and marketing [4], [5].

*Corresponding author: [email protected]

2021 UTHM Publisher. All rights reserved.
86
penerbit.uthm.edu.my/ojs/index.php/jscdm
Hasan et al., Journal of Soft Computing and Data Mining Vol. 2 No. 1 (2021) p. 86-97

Recognition is a modern field of study that has piqued the interest of researchers since it becomes simple to use by
using OpenCV-based Python. Face recognition technologies have a variety of applications in public protection,
entertainment, man-machine contact, and social networking, such as Facebook's automated tag recommendation on
images. It's also been seen in educational and non-educational institutes' attendance control, financial offices, voter
registration, and other areas [6], [7]
In this paper, we emphasize the important role of OpenCV in face detection and face recognition, what algorithm
can be used in OpenCV for face detection and face recognition, then state the OpenCV modules and explain OpenCV
based on Python and mention the applications for OpenCV are. Finally, we assessment and compared recent literature
reviews that use OpenCV to detect and recognize the human face in a variety of fields in order to improve human life.
The rest of this paper is organized as follows: in section 2, face detection is described. In section 3, face
recognition is explained. In section 4, OpenCV library and OpenCV algorithm are explained. In section 5 , Modules of
OpenCV are explained. In section 6, OpenCV based on Python is described. In section 7, the assessment of literature
reviews and comparison table are discussed. Section 8 concludes this paper.

2. Face Detection
Face detection has received much attention in recent years because of its applications in computer and human
interaction. Face detection is a subset of image processing. Image processing is primarily a technique for compressing,
improving, or extracting valuable information from images. Facial recognition technology can identify single or many
faces in a picture, removing unwanted background noise. A face identification algorithm must basically categorize
pictures into two groups based on whether or not they include a face. The face detection algorithm's goal is to
thoroughly examine the picture, identify the existence of faces in the image, and remove the background from the
image. Face detection mistakes are classified into two types: false negative and false positive. A false positive occurs
when a face is identified in a picture that does not include any faces. A false negative occurs when the algorithm rejects
the existence of anything in the picture. The detection rate is the ratio of the number of faces identified by humans to
the number of faces detected properly by the system. The detection rate of the face detection algorithm should be as
high as possible[1], [8].

3. Face Recognition
Facial recognition is the world's most advanced and quick biometric technology. It takes advantage of the most
visible human body component, the face, in a non-intrusive way. According to worldwide data, most individuals are
unaware of the face recognition process that is taking place on them, making it one of the least invasive procedures
with the least amount of delay. The facial recognition algorithm examines the many features of a face in the input
picture. This biometric has been extensively, and maybe exaggeratedly, lauded as a great method for identifying
possible dangers such as terrorists, scam artists, and so on, but it has yet to gain widespread acceptance in high-level
use. Biometric face recognition technology is expected to surpass fingerprint biometrics as the most common method of
user identification and authentication in the near future [8], [9], [10].

4. OpenCV Library
It is a massive open-source image processing, machine learning, and computer vision library. OpenCV is
compatible with a broad range of programming languages, including Python, C++, and Java. It will analyze
photographs and videos to recognize artifacts, faces, and even human handwriting. When paired with many other
libraries, like Numpy, a high-performance library for turning machines, achieve a good performance; that is, all
services that can be performed in Numpy can also be integrated with OpenCV.It is written based on C++ and has a C++
interface as its main interface, but it also has a less robust but still detailed older Language training. Both the latest
technologies and algorithms are visible in the C++ GUI. Python, Java, and MATLAB/OCTAVE bindings are available
[3]. Wrappers in a variety of programming languages have been created to promote broader acceptance. JavaScript
plugins for a variant of OpenCV functions are published as OpenCV.js in version 3.4, which can be used on web
platforms. The OpenCV project, which was officially unveiled in 1999, was originally Intel's research program to
support CPU-intensive applications [11].OpenCV is a popular platform for implementing face detection and
recognition algorithms. The following are some often used OpenCV algorithms.

4.1 Haar Cascade

Haar Cascade is an effective method for detecting objects. It's a machine-learning-based method in which a
cascade of actions is learned from a large number of positive and negative images. It becomes used to seeing things in
different frames [12]. Fig. 1 shows the view of the Haar cascade classifier. Fig. 2 shows the Haar cascade flowchart.

87
 Hasan et al., Journal of Soft Computing and Data Mining Vol. 2 No. 1 (2021) p. 86-97

Fig. 1 - View of Haar cascade classifier

Fig. 2 - Haar Cascade flowchart [1]

4.2 LBP (Local Binary Pattern)

It is a simple but effective texture operator that labels pixels in an image by thresholding the pixels' neighborhood
and treating the result as a binary number. The LBP texture operator has become a common approach in a variety of
applications due to its discriminative power and computational simplicity. It can be viewed as a unifying solution to
texture analysis's historically divergent statistical and structural models. The LBP operator's robustness to monotonic
grayscale changes induced, for example, by illumination variations is perhaps its most significant property in real-world
applications. Another key feature is its computational simplicity, which allows it to analyze images in difficult real-
time scenarios [13]. Fig. 3 shows the description of facial expressions with local binary patterns.

88
Hasan et al., Journal of Soft Computing and Data Mining Vol. 2 No. 1 (2021) p. 86-97

Fig. 3 - Description of facial expressions with local binary patterns [13]

4.3 EigenFaces
It is a method that employs PCA (Principal Component Analysis) to minimize dimensionality and find the
strongest vectors for distributing facial images onto existing facial spaces. The primary goal of PCA is to identify the
best vectors to explain the distribution of facial images in picture space into face space. According to the eigenvalue
distribution, m eigenvector is used to construct the principal component amount. Eigenvector and eigenvalue are
calculated from the qualified facial image's covariance matrix. The eigenvector is sorted by eigenvalue (high to low)
and M first eigenvectors are chosen to form the principal variable [1], [14].

4.4 FisherFaces
It is a face recognition system that several researchers have demonstrated to identify faces accurately. FisherFace is
a calculation model that combines the PCA (Principal Component Analysis) calculation model with Fisher's Linear
Discriminant (FLD). PCA is used to minimize input data in order to simplify and accelerate the FLD operation. On the
other hand, FLD is used to generate a distribution matrix to aid in classification and identification. A series of
FisherFaces is generated using the PCA and FLD calculation models. This facial recognition process consists of four
major steps: face identification, PCA estimation, calculation, and classification [15].

4.5 LBPH (Local Binary Pattern Histogram)

LBP is a highly effective texture operator. That compares each neighboring pixel's threshold value to the value of
the center pixel. It takes into account outcomes in the context of binary numbers. LBP is a common technique in a
variety of applications due to its discriminative strength and simplicity. LBP was identified for the first time in 1994.
Since then, it seems to have evolved into a more efficient texture classification algorithm. It was later discovered that
combining LBP with histograms of directed gradient descriptors increases its accuracy on the same dataset. LBP has
additional capabilities such as monotonic grey-scale improvements and statistical simplicity, allowing it to interpret
images in real-time applications [1]. Fig. 4 shows the LBPH algorithm for face recognition.

Fig. 4 - LBPH Algorithm for Face Recognition [1]

89
 Hasan et al., Journal of Soft Computing and Data Mining Vol. 2 No. 1 (2021) p. 86-97

4.6 YOLO
Yolo is an abbreviation for (you only look once). It is the most recent real-time object detection system that uses a
single neural network to process the entire image [16]. This network splits the picture into sections and estimates the
bounding boxes and probabilities for each [17]. The estimated probabilities are used to weigh these bounding boxes.
The testing phase examines the whole picture, so the image's global meaning guides its predictions. Can detect objects
form videos or image [18] [19]. Fig. 5 shows the Yolo process.

Fig. 5 - The Yolo process [18]

4.7 Faster R-CNN

It is a popular object detection architecture proposed in 2015 by Ross_Girshick and is one of the most well-known
object detection architectures that employ convolution neural networks. The implementation of the Region Proposal
Network enables (Faster R-CNN) easier and quicker (RPN). RPN is a completely convolutional network that has been
trained side-to-side, and forecasts object boundaries and object ratings at each detection. Since RPN is so critical to
(Faster-R-CNN), and remains one of the strongest entity detection frameworks open to researchers, most of this piece
would concentrate on RPN architecture and the notions of anchor boxes and suppression non-maximum [18]. Fig. 6
shows the Faster_ R-CNN step.

Fig. 6 - Faster_ R-CNN step [18]

90
Hasan et al., Journal of Soft Computing and Data Mining Vol. 2 No. 1 (2021) p. 86-97

4.8 Single Shot Detectors (SSDs)

The SSD method is focused on the feed-forward convolutional network that generates a permanent border-box
array and results in the existence of class-based entity instances in these boxes and a non-maximum deletion stage to
generate final detection. The early network layers are built on a common image classification design [16], [20]. Fig. 7
shows the SSD.

Fig. 7 - SSD [16]

5. Modules of OpenCV
5.1 Main Modules
Main modules are the main components of OpenCV and are included in the package versions by necessity. They
are called core modules since they include essential functions, including image recognition, filtering, and
transformation [8], [21].

5.2 Extra Modules

Extra modules are not included in the OpenCV release by default. Additional machine vision functionalities, such
as text detection, are applied to these modules [21].
The main modules are described below:
 Core: Have most of OpenCV's core features.
 Imgproc: Picture processing tools such as transforms, manipulations, and filtering are used.
 Imgcodecs: Image reading/writing features are included.
 Videoio: Video reading/writing features are included.
 Highgui: Used for creating GUI to display output.
 Video: Motion detection and monitoring.
 calib3d: This package includes calibration and 3D reconstruction functions for estimating translation between
multiple pictures.
 features2d: Item identification and categorization algorithms utilizing keypoint detection and descriptor
extraction algorithms, which are included in this library.
 Objdetect: used for detecting object.
 Dnn: Classify and detecting objects, between other objects.
 Ml: Used for regression and classification, and it encompasses the vast majority of machine learning.
 Flann: Supports optimized algorithms for searching for high-dimensional attributes in massive data sets using
nearest neighbor search. Fast Library for Approximate Nearest Neighbors (FLANN) is an acronym for Quick
Library for Estimated Nearest Neighbors (FLANN)[22]. Photo: Removes noise and creates HD images,
among other roles for photography-related computer vision.
 Stitching: Used for stitching image.
 Shape: Deal with issues such as shape transformation, pairing, and distance.
 Superres: Deal with enhancement and resolution algorithms.
 Videostab: Have algorithms for stabilizing video.
 Viz: create 3D display window for widgets.

6. OpenCV based on Python

Python was created by Guido van Rossum, who focused on writing clear, easy to understand code. With shorter
code, the programmer will articulate the same ideas in fewer characters. Python is slower than some programming
languages like C/C++. C/C++ can be expanded with Python effectively, but one other valuable aspect of Python is that
it is generalizable and flexible. This function assists in the implementation of C++ codes that we can then transform

91
 Hasan et al., Journal of Soft Computing and Data Mining Vol. 2 No. 1 (2021) p. 86-97

into Python modules for ease of usage[3]. These two things offer great benefits: first, the code is as quick as C++ code
(as it is), and second, Python is simple to program in. The Python bindings for OpenCV function like this. The Python
implementation was built on top of the initial C++ code. Numpy's contribution helps here. Numpy is a highly structured
library that often offers Python-level statistical capabilities. It does things in a MATLAB-like fashion. Conversion to
Numpy arrays is done, and array processing is done on all the OpenCV structures. This is very good because whatever
functions it can do in Numpy, it can now combine with OpenCV, making you more efficient. SciPy, which also
supports NumPy, may be used along with this. So Python with OpenCV offers a good way to build out computer vision
projects [11] quickly.

7. Assessment of Literature Reviews

OpenCV is a library for image and video processing that is used for image and video analysis, among other things.
We review some recent literature concerning Face Detection and Recognition Using OpenCV.

7.1 Face Detection

Alcantara et al. [22] proposed a system for tracking and detecting the human head in real-time video using the
OpenCV library. The proposed system will employ a Haar-like classifier to detect the head, Haar Training to train the
system, and use CMT object tracking algorithm to track the head, where the result show 68% accuracy of CMT to
tracking the person's head and 83% accuracy in detecting the person's head.
Gupta [7] proposes a method for detecting emotions in real-time and fixed pictures. So, before attempting emotion
detection, they must first identify faces in static photographs or real-time videos using the Haar classifier of OpenCV. If
the face has been identified, it may be cropped and analyzed to detect additional facial landmarks. The datasets are then
conditioned using facial landmarks and categorized according to the eight emotions using the SVM, a machine learning
algorithm. They obtained an accuracy of about 93.7 percent using SVM. These facial landmarks may be tweaked to
improve precision.
Lee et al.'s [4] research aims to solve the classic issue of face detection in various lighting conditions and create an
intelligent and effective human face detection system using Visual Studio 2015 software framework and OpenCV
technologies. They demonstrated by experimentation that the photo processing approach applied in their paper
completed facial recognition under various lighting conditions, which really is a significant advancement in face
recognition technology.
Gupta et al. [16] proposed a system to improve universities' traditional attendance systems and avoid time wasted
in counting traditional attendance based on image processing methods. The Student Attendance layout structure's
primary function is to perform, incorporate, and manage attendance notes for a pupil, perform an automatic estimate on
the amount of presented and absentees depending on the topic and affability of the class, and then produce an
automated document or spreadsheet. They used OpenCV library, Haar-Cascade for face detection, and LBPH for face
recognition; after that, individual student training occurred, and eventually, the device produced a spreadsheet that
provided the number of students present in the classroom with a picture or video captured live.
Das et al. [19] offer a simpler approach to Face-Mask Detection by using several fundamental Machine Learning
packages such as Scikit, OpenCV, TensorFlow, and Keras. The suggested procedure accurately senses the face in the
picture and then determines whether or not it has a mask on it. It may track a face and a mask in motion as a
surveillance mission artist. On two separate datasets, the system achieves a precision of up to 95.77 percent and 94.58
percent, respectively. They investigate optimized parameter values for the Serial CNN model to properly detect the
existence of masks without triggering over-fitting.
Hoque et al. [5] designed software that can identify people's faces or faces from live video streaming. ATMega328p
Arduino Uno-based Micro Controller with Pan-Tilt function and OpenCV are at the heart of the control scheme. The
hair–Cascade, Camshift, Housedorff gap, AdaBoost, Viola Jones, etc. are all used to identify human faces. They used
the Cascade algorithm Haar Classifier to classify faces.
Mehariya et al. [20] designed a system to overcome the problem of students not attending lectures in classrooms in
universities or any place requiring workers' attendance. As non-attendance is a waste of the infrastructure, manual
absence management is a tedious task and a waste of time, as the traditional absence counting system may be cheated.
They used OpenCV to find the number of students in the class and built an algorithm giving the best occupancy ratio.
The method suggested would be different from the conventional approach. The detect Multiscales procedure is used to
count the number of students in a class, and the occupancy ratio is used to assign a complex classroom. The first way to
classify the individual in a picture is the Multiscale detect method. It does not take items into consideration. This count
is saved in the forecasting firebase, and the occupancy ratio is used for complex classroom distribution. This reduces
the waste of room and allows us to prepare the schedule effectively. Factors such as classroom scale, projectors in
classrooms, and dynamic allotment capability are taken into account. Google's cloud database (Firebase) is used to
maintain and store college/university files. The personal records, rosters, schedules, and attendance sheets are included.
Sriratana et al. [11] created a Personal Identifier scheme by combining the Viola and Jones algorithms with the
OpenCV library, and Python coded on the Raspberry Pi controller board. The framework successfully demonstrates all

92
Hasan et al., Journal of Soft Computing and Data Mining Vol. 2 No. 1 (2021) p. 86-97

planned goal points, including cost-effective deployment, simple installation, and real-time detection. Just 8-9 percent
of errors were discovered during tests of 150 samples, demonstrating the system's high precision and performance.
Patel et al. [23] suggested a method that senses whether or not the driver is asleep when operating a car or other big
vehicle, and if the driver is suspected sleepy, the system will alert the driver to wake up and stop driving. One of the
reasons for public road collisions is the driver's drowsiness when driving. It is necessary to develop an efficient
approach to predict somnolence as soon as the driver becomes drowsy. This may help to reduce the large number of
injuries that occur. This method would aid in the reduction of drivers' sleeping injuries. In the OpenCV setting, he used
real-time image processing with a vision device and the technique of facial expression and eye blinking.

7.2 Face Recognition

Boyko et al. [21] compared the efficiency of two main libraries in computer vision (Dlib and OpenCV) and
designed two basic face recognition systems for the two libraries. Based on his results, he demonstrates that the
OpenCV library is more efficient and has improved face detection and recognition results than the Dlib library. It also
implies that OpenCV is best suited for developing recognition software for the IoT framework.
Sarkar and Sikka [9] investigate and evaluates various classifiers used in facial embedding classification. They also
focus on a Python-based face recognition pipeline that can be used to build a face recognition framework on compact
low-power hardware devices. The technique mentioned employs pertained models and structures, resulting in cutting-
edge efficiency without the need for strong hardware. On the LFW dataset, the suggested approach reaches an accuracy
value of 99.4.
Sharma [18] suggested a facial recognition system for specialized applications, including entry and protection,
purchases, and criminal identity. The identification method will be focused on facial recognition, which will be face
detection, feature extraction and classification, and real-time face recognition. They used Haar-like for face detection
and LBPH for face recognition, all this was done by using OpenCV in the Python environment. Kivy is used to build
user interfaces to render the proposed system executable through several platforms.
James and Nettikadan [12] introduced a system for monitoring in real-time inside school buses. By utilizing image
processing, using an image to identify a student by a camera. This machine tracks the videos inside the bus and
recognizes the students and their gestures. The proposed system recognizes the students' faces and keeps a count of
their numbers. If it is absolutely necessary, the system may even sound like a warning to draw the public's attention by
using technologies from OpenCV and implementing them in Python. Face detection was performed using the Haar-
Cascades classifier, and face recognition was performed using Eigenfaces and Lbph. This removes the majority of the
disadvantages of manual attendance systems, such as the simple manipulation of attendance records.
Balachandran et al. [13] designed an efficient application for facial recognition using AI. For the neural network,
the VGGFace framework was used. The application is divided into two phases: training and recognition. The training
phase involves adding new faces to the scheme, while the identifying phase involves determining a face's identification.
The program will operate on several cores with no problems.
Apoorva. et al. [15] proposes a technique for robust face recognition in a real-time setting. Haar-cascade is one of
the facial recognition algorithms. On the OpenCV website, they tracked faces using Haar-like classifiers. Face
detection has a good level of precision. Since the computation period is concise, the proposed method will successfully
identify more than one face that is helpful for rapidly looking for suspects.
Srivastava et al. [17] created an Attendance System by combining facial recognition technologies with (OpenCV)
algorithm. By simply maintaining a record of clock-in and clock-out times, this program can simplify the attendance
automation process and help faculties to access students' details.
Soomro et al. [6] built a standalone authentication program using a facial recognition method with (NI VISION,
LabVIEW, NI MyRIO, OpenCV). The developed framework should identify and enable authenticated individuals to
bear in mind internal system weaknesses in a real-time context, especially those induced by a pause. The entire
framework is made up of hardware and software, with parallel processing methods and modules like the NI MyRIO
FPGA being included.
Sharma et al. [24] propose a system of recognition that a blind individual may benefit from. This paper applied a
hand gesture identification method and facial recognition system to conduct many tasks. Dynamic images are from a
dynamic video and are interpreted by algorithms. In the Hand Motion scheme, skin color recognition was carried out in
YCbCr colors and a convex hand-based defect character point was used to identify various characteristics, such as
fingertips and angles between fingers. Different activities such as spinning the fan or lights may be carried out
according to the recognized gesture. OpenCV, Haar-Cascade, and LBPH are respectively used to detect the mask and to
identify the face.
Salihbasic and Orehovacki [14] outline and describe the whole method of creating an Android app to recognize a
person's gender, age, and face in depth. Face detection and recognition techniques and development software used in
the development of an Android mobile application is defined and discussed. The software solution explains how to use
the OpenCV library and displays the actual results of the smartphone app using photos.
Zhu and Cheng [25] present. An efficient Attitude Tracking Algorithm (EATA) based on OpenCV is used for
facial recognition in the Intelligent door lock system. A specialized application has been created to allow the operation

93
 Hasan et al., Journal of Soft Computing and Data Mining Vol. 2 No. 1 (2021) p. 86-97

of devices as well as the display of the location. An automated door surveillance system using raspberry pi python,
USB camera, and OpenCV is given for monitoring and security reasons. The system status database was created at a
certain point in time. The gadget is very cheap, simple to set up, and simple to use. The Attitude Tracking Algorithm is
extremely accurate and performs well.

Table 1 - Comparison table

Ref. Aim classifier accuracy Result/purpose
[21] SVM OpenCV more 83% head The OpenCV library is more productive,
2018 accurate than dlib detect has improved facial recognition and
detection accuracy
[22] Head Haar-like 68% tracking The proposed system successfully detected
2018 Detection and CMT the head of a human using OpenCV
Tracking libraries, specifically using Haar-like
Cascade attribute detection.
[9] comparative MLP 99.1% The results show that Logistic Regression
2018 of classifying Extra Tree 86.4% outperforms the other algorithms for face
the face using classification in terms of speed and
different (Random 95.9% accuracy.
classifiers Forest)
KNN 99.3%
RadialSVM 98.4%
GaussianNB 98.8%
LinearSVM 99.2%
(Logistic
Regression)
99.4%
[7] real-time Haar 93.7% The findings suggest that with today's
2018 recognition SVM computing power, user-independent,
Facial completely automated real-time coding of
emotion facial expressions in a continuous video
stream is a goal that can be achieved.
[4] Face Haar 80% The experiment demonstrated that the
2018 Detection picture processing system has facial
under recognition in various lighting conditions.
Different
Lighting
[18] Designing of Haar-like 80% The system is tested by more than 150
2019 Face LBPH people and has a reliability of approximately
Recognition 80%. It is measured with multiple cameras
System in various settings, and lighting conditions,
and the findings are about the same. A
Logitech C90 USB webcam is used here.
[12] Student Haar- 85% The system watches the bus and detects the
2019 Monitoring Cascades students and their movements,
System for LBPH acknowledges the faces of the students, and
School Bus their count is also tracked and alerts the
Eigenfaces audience if necessary.

[13] Face Keras VGG- module The results collected were rather similar to
2019 Recognition Face performance what was predicted. The efficiency gained
in Parallel depends on by simply running the program on a
Computer the number machine with more computing capacity and
of processors cores than a simple laptop was important.
[14] Recognition cascade successfully Gender, face, and age recognition was
2019 of the gender, LBP work but achieved
age, and face affected by
of the person LBPH mobile type,
face
coverage,
expressions
of face,

94
Hasan et al., Journal of Soft Computing and Data Mining Vol. 2 No. 1 (2021) p. 86-97

person pose
[15] criminal Haar-like 80% Since the computation period is very short,
2019 identification the proposed method will successfully
by face identify more than one face, which is helpful
recognition for rapidly looking for suspects.
automated
[16] Automated KNN 97% Prevent students from marking fake
2020 Attendance HaarCascade attendance for other students. It would also
System save faculty resources in universities by
LBPH eliminating the need for them to take
attendance of students who are present in
class, and it will be able to send monthly
attendance reports to students' parents by
email.
[17] Attendance HaarCascade 95% The results explicitly demonstrate that as the
2020 System in Paul – Viola. facial angle increases, face identification and
real-time recognition rate decreases.
[6] real-time Haar-cascade 90% The primary goal of this project is to
2020 system of camshift completely concentrate on the electronic
electronic voting mechanism and the protection of
voting every organization.
authorized by
face
recognition
[19] Face Mask Cascade between Since wearing a mask which becomes
2020 Detection to CNN 95.77%, mandatory before the Covid-19 crisis is
protect from 94.58% resolved, the implemented model can make
Covid-19 a significant contribution to the public health
care system.
[5] Autonomous Haar Cascade 83% Recognize human faces with some kind of
2020 Face camera and issue an alarm with a buzzer and
Detection an automatic-on light bulb that makes it
System from noticeable from a long distance.
Real-time
Video
Streaming
[20] Counting Cascade 90 to 100%. By installing cameras in the hallway, the
2020 students in model may be used to detect students who
classroom are skipping classes.
allocation
[24] A Face Haar hand gesture The developed system will function as a
2019 Recognition Cascade recognized is virtual assistant for a blind individual using
and Static 95.2% hand gestures and face recognition.
Hand Gesture LBPH
Facial
System for recognition is
the Blind 92%.
[11] Personal Cascade 90% Just 8-9 The proposed application helps companies
2018 Identifier Haar percent of calculate work attendance and detect cases
application errors were of fraud in work attendance compared to old
discovered work attendance monitoring methods.
after
analyzing
150 samples.
[25] door lock Attitude 95% According to the testing findings, the
2020 intelligence Tracking suggested system is more efficient, uses less
based on face Algorithm power, and is more cost-effective.
recognition (ATA)
[23] Drivers' EEG 90% This method would aid in the reduction of
2018 Somnolence ANN drivers' sleeping injuries. In the OpenCV
Detection setting, he used real-time image processing
method in with a vision device and the technique of
Real Time facial expression and eye blinking.

95
 Hasan et al., Journal of Soft Computing and Data Mining Vol. 2 No. 1 (2021) p. 86-97

Based on Table 1, we compared the accuracy of techniques and classifiers of OpenCV used in a number of the
studies reviewed of face detection and face recognition in different areas of computer vision. From our observations of
the 20 reviewed studies (publications) (2018–2020) discussing the used methods of face detection and recognition in
the field of computer vision using OpenCV. It was found that OpenCV can be used in different fields like face
detection, recognition the faces, and Recognize facial expressions [26], [27]. Also, it was found that we can use
OpenCV in the security field like criminal identification [28]. Additionally, others used OpenCV to automate the
attendance of students or workers in institutions. At the same time, other researchers used OpenCV to reducing traffic
accidents on the roads.

8. Conclusion
Computer Vision is the subfield of Artificial Intelligence, where computers are trained to process the image and
extract the important features from the images or videos. Open Computer Vision (OpenCV), a python library written in
C++, provides various functionalities for computer vision applications. Applications of computer vision are object
detection, face recognition, medical diagnosis, etc. In this paper, we emphasize the important role of OpenCV in face
detection and face recognition. We illustrate the popular algorithms in OpenCV that are used for face detection and face
recognition. Then state the OpenCV modules and explain OpenCV based on Python and mention the applications for
OpenCV. Finally, we assessment and compared recent literature reviews that use OpenCV to detect and recognize the
human face in a variety of fields in order to improve human life.

Acknowledgement
The authors would like to thank the IT Department, Technical College of Informatics Akre, Duhok Polytechnic
University, Duhok, Kurdistan Region, Iraq and College of Engineering, Nawroz University, Duhok, Kurdistan Region,
Iraq.

References
[1] Jagtap, A. M., Kangale, V., Unune, K., & Gosavi, P. (2019, February). A Study of LBPH, Eigenface, Fisherface
and Haar-like features for Face recognition using OpenCV. In 2019 International Conference on Intelligent
Sustainable Systems (ICISS) (pp. 219-224). IEEE
[2] Sigut, J., Castro, M., Arnay, R., & Sigut, M. (2020). OpenCV basics: a mobile application to support the teaching
of computer vision concepts. IEEE Transactions on Education, 63(4), 328-335
[3] Adusumalli, H., Kalyani, D., Sri, R. K., Pratapteja, M., & Rao, P. P. (2021, February). Face Mask Detection Using
OpenCV. In 2021 Third International Conference on Intelligent Communication Technologies and Virtual Mobile
Networks (ICICV) (pp. 1304-1309). IEEE
[4] Mostafa, S. A., Mustapha, A., Gunasekaran, S. S., Ahmad, M. S., Mohammed, M. A., Parwekar, P., & Kadry, S.
(2021). An agent architecture for autonomous UAV flight control in object classification and recognition
missions. Soft Computing, 1-14
[5] Hoque, M. A., Islam, T., Ahmed, T., & Amin, A. (2020, March). Autonomous face detection system from real-
time video streaming for ensuring the intelligence security system. In 2020 6th International Conference on
Advanced Computing and Communication Systems (ICACCS) (pp. 261-265). IEEE
[6] Soomro, Z. A., Memon, T. D., Naz, F., & Ali, A. (2020, January). FPGA Based Real-Time Face Authorization
System for Electronic Voting System. In 2020 3rd International Conference on Computing, Mathematics and
Engineering Technologies (iCoMET) (pp. 1-6). IEEE
[7] Gupta, S. (2018, January). Facial emotion recognition in real-time and static images. In 2018 2nd international
conference on inventive systems and control (ICISC) (pp. 553-560). IEEE
[8] Dino, H., Abdulrazzaq, M. B., Zeebaree, S. R., Sallow, A. B., Zebari, R. R., Shukur, H. M., & Haji, L. M. (2020).
Facial expression recognition based on hybrid feature extraction techniques with different classifiers. TEST
Engineering & Management, 83, 22319-22329
[9] Kashinath, S. A., Mostafa, S. A., Mustapha, A., Mahdin, H., Lim, D., Mahmoud, M. A., ... & Yang, T. J. (2021).
Review of Data Fusion Methods for Real-Time and Multi-Sensor Traffic Flow Analysis. IEEE Access
[10] Abdulrazaq, M. B., Mahmood, M. R., Zeebaree, S. R., Abdulwahab, M. H., Zebari, R. R., & Sallow, A. B. (2021,
February). An analytical appraisal for supervised classifiers’ performance on facial expression recognition based
on relief-F feature selection. In Journal of Physics: Conference Series (Vol. 1804, No. 1, p. 012055). IOP
Publishing
[11] Sriratana, W., Mukma, S., Tammarugwattana, N., & Sirisantisamrid, K. (2018, July). Application of the OpenCV-
Python for Personal Identifier Statement. In 2018 International Conference on Engineering, Applied Sciences, and
Technology (ICEAST) (pp. 1-4). IEEE
[12] James, C., & Nettikadan, D. (2019, April). Student monitoring system for school bus using facial recognition.
In 2019 3rd International Conference on Trends in Electronics and Informatics (ICOEI) (pp. 659-663). IEEE

96
 Hasan et al., Journal of Soft Computing and Data Mining Vol. 2 No. 1 (2021) p. 86-97

[13] Balachandran, B., Saad, K. F., Patel, K., & Mekhiel, N. (2019, December). Parallel Computer for Face
Recognition Using Artificial Intelligence. In 2019 14th International Conference on Computer Engineering and
Systems (ICCES) (pp. 158-162). IEEE
[14] Salihbašić, A., & Orehovački, T. (2019, May). Development of android application for gender, age and face
recognition using opencv. In 2019 42nd International Convention on Information and Communication
Technology, Electronics and Microelectronics (MIPRO) (pp. 1635-1640). IEEE
[15] Apoorva, P., Impana, H. C., Siri, S. L., Varshitha, M. R., & Ramesh, B. (2019, March). Automated criminal
identification by face recognition using open computer vision classifiers. In 2019 3rd International Conference on
Computing Methodologies and Communication (ICCMC) (pp. 775-778). IEEE
[16] Gupta, N., Sharma, P., Deep, V., & Shukla, V. K. (2020, June). Automated attendance system using OpenCV.
In 2020 8th International Conference on Reliability, Infocom Technologies and Optimization (Trends and Future
Directions)(ICRITO) (pp. 1226-1230). IEEE
[17] Srivastava, M., Kumar, A., Dixit, A., & Kumar, A. (2020, February). Real time attendance system using face
recognition technique. In 2020 International Conference on Power Electronics & IoT Applications in Renewable
Energy and its Control (PARC) (pp. 370-373). IEEE
[18] Sharma, V. K. (2019, May). Designing of Face Recognition System. In 2019 International Conference on
Intelligent Computing and Control Systems (ICCS) (pp. 459-461). IEEE
[19] Das, A., Ansari, M. W., & Basak, R. (2020, December). Covid-19 Face Mask Detection Using TensorFlow, Keras
and OpenCV. In 2020 IEEE 17th India Council International Conference (INDICON) (pp. 1-5). IEEE
[20] Mehariya, J., Gupta, C., Pai, N., Koul, S., & Gadakh, P. (2020, July). Counting Students using OpenCV and
Integration with Firebase for Classroom Allocation. In 2020 International Conference on Electronics and
Sustainable Communication Systems (ICESC) (pp. 624-629). IEEE
[21] Boyko, N., Basystiuk, O., & Shakhovska, N. (2018, August). Performance evaluation and comparison of software
for face recognition, based on dlib and opencv library. In 2018 IEEE Second International Conference on Data
Stream Mining & Processing (DSMP) (pp. 478-482). IEEE
[22] Alcantara, G. K. L., Evangelista, I. D. J., Malinao, J. V. B., Ong, O. B., Rivera, R. S. D., & Ambata, E. L. U.
(2018). Head detection and tracking using OpenCV. In 2018 IEEE 10th International Conference on Humanoid,
Nanotechnology, Information Technology, Communication and Control, Environment and Management
(HNICEM) (pp. 1-5). IEEE
[23] Patel, R., Patel, M., & Patel, J. (2018, April). Real Time Somnolence Detection System In OpenCV Environment
for Drivers. In 2018 Second International Conference on Inventive Communication and Computational
Technologies (ICICCT) (pp. 407-410). IEEE
[24] Sharma, S., & Jain, S. (2019, March). A static hand gesture and face recognition system for blind people. In 2019
6th International Conference on Signal Processing and Integrated Networks (SPIN) (pp. 534-539). IEEE.
[25] Zhu, Z., & Cheng, Y. (2020). Application of attitude tracking algorithm for face recognition based on OpenCV in
the intelligent door lock. Computer Communications, 154, 390-397
[26] Chandan, G., Jain, A., & Jain, H. (2018, July). Real time object detection and tracking using Deep Learning and
OpenCV. In 2018 international conference on inventive research in computing applications (ICIRCA) (pp. 1305-
1308). IEEE
[27] Khan, M., Chakraborty, S., Astya, R., & Khepra, S. (2019, October). Face Detection and Recognition Using
OpenCV. In 2019 International Conference on Computing, Communication, and Intelligent Systems
(ICCCIS) (pp. 116-119). IEEE
[28] Palekar, R. R., Parab, S. U., Parikh, D. P., & Kamble, V. N. (2017, April). Real time license plate detection using
openCV and tesseract. In 2017 international conference on communication and signal processing (ICCSP) (pp.
2111-2115). IEEE

97

View publication stats