www.ijcrt.

org © 2021 IJCRT | Volume 9, Issue 7 July 2021 | ISSN: 2320-2882

FACIAL RECOGNITION SYSTEM USING

OPENCV
Vivek Anand1, Vimal Singh Parihar2, Shubham Kumar Sharma3, Vikas Singhal4 and Pramod Kumar
Sethy5
1
Department of Computer Science and Engineering, Krishna Engineering College, India
2
Department of Computer Science and Engineering, Krishna Engineering College, India
3
Department of Computer Science and Engineering, Krishna Engineering College, India
4
Department of Computer Science and Engineering, Krishna Engineering College, India
5
Department of Computer Science and Engineering, Krishna Engineering College, India

Abstract
The main objective behind the facial recognition system is the computer science. In this letter, we have taken our research
certitude that every person has a unique face. As we know that every using Open CV. Reasons for using OpenCV are discussed
person has unique fingerprint, similarly every individual face have further in this paper.
unique features. Here we use features of face of an individual. We
can stored the features of the faces of many individuals and they can
be identified according to their face features. Facial testimony and 2. REASONS FOR USING OPENCV
facial recognition are difficult and challenging piece of work. For
facial recognition systems to be authentic, they must work accurately (a) Speed- OpenCV uses C/C++ library functions which provide
and precisely. The facial recognition technique captured the image the computer with machine language code and help in faster
using the camera and mapping it for comparison to the images stored execution, use of OpenCV results in more use of time and
in the database. If the captured image is matched with any of the resources in image processing and less in explaining.
stored images then it shows face matched otherwise it shows face not
matched. This paper elaborate in detail the entire process of facial
recognition system using OpenCV library. We use Haar cascading
(b) Portability- As OpenCV implements on C, therefore any
algorithm using OpenCV library for the face detection. devices which run on C can run OpenCV. It can work well with
Windows or Linux.
Keywords: Face detection, face recognition, Haar cascade, OpenCV.
(c) Cost- OpenCV is free for all because it is a BSD license so it
is free of cost.
1. INTRODUCTION
3. RELATED WORK
In the last two decades, the facial recognition system has become
one of the most important and interesting research areas. A facial
There are five basic steps involves in a facial recognition system
recognition system is a software application for certifying an
includes image capture, face detection, feature extraction,
individual and recognizing him/her with images or videos from a
comparison, and face recognition. Face recognition technology
source. Facial recognition can be done speedily and accurately
analyzes the unique shape, pattern, and positioning of facial
with the open-source platform called OpenCV. A path from a
features. Face recognition is a very complex technology and is
face and a picture database are favorite facial features. It is
largely software-based.
usually compared to biometrics such as fingerprints and eye
It holds the record for identifying the captured image. Various
investigation systems, and security systems and used in thumb
features like nose shape, eye shape, lips, skin color, skin tone,
detection systems. The OpenCV library makes programming
etc. The face recognition system first captures an image from a
easy to use. This comes up with advanced proficiencies like face
camera as input and finds the face in the image for face
detection, face tracking, facial recognition, and many more
detection. Face extraction includes obtaining the features from
methods for artificial intelligence (AI). The main advantage of
the face captured by the camera. Then it compares the features
the OpenCV library is, it is a multi-platform framework; it
with the features of images stored in the database and according
supports Windows, Mac OS, Mac OS X.
to which it gives result. If the features are matched with stored
Its challenging work includes face recognition on the lowest
image features then it identified the person otherwise not
computing cost framework such as smartphones and embedded
identified. However facial recognition system still has few
devices. For the person's testimony, facial recognition is used.
drawbacks as it cannot detect the face due to overlapping of
Everyone has unique features that do not share with another
faces or difficult recognition of two faces having the same
person.
features.
Currently, there are many devices and application which use
face detection technology to recognize and detect a face such as
Facebook. Therefore, face detection is not new in the vision of

IJCRT2107697 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org g446

www.ijcrt.org © 2021 IJCRT | Volume 9, Issue 7 July 2021 | ISSN: 2320-2882
C. IMAGES HAVING A COMPLEX BACKGROUND

Face detection in the complex background can be done using the

MUHULANOBIC metric. It is based on the detection of human
faces in two-dimensional natural images. It discretely makes use
of the process of a color fraction of the image that is being taken
as the input. This fraction of the color is being performed by sill
the image in the tone color space. It is taken care of the effects of
distinguishing the color in the human skin when the lighting has
been changed in the image. Then the results of the image are
collected together for any other examination. At the end the
difference between the faces and the remaining complex
background, a multi-layer perceptron neural network is used
with the invariant moments as an input factor.

3.2 FINDING FACES VIA MOTION

3.2.1 USE OF BLINK DETECTION TECHNIQUES

Fig 1. Steps of Facial Recognition System
Blinking is a reflexive act conducted by humans. This is a very
hurry process. Some humans might not even take care of
3. FINDING FACES blinking in daily life but blinking as a process has been proved
to detect the presence of a human significantly at any frame of
Finding faces is the most essential part of face detection. There time. Blinking provides a casual time and space signal which is
are different techniques by which faces can be found. In this unique to every other individual. Therefore, the blinking process
paper, we will compare the various algorithms used before and can identically act as a biometric means of measurement to
analyzing them. Face detection is the most important step of a detect the presence. An algorithm is used to make blinking make
facial recognition system still the technique and algorithm used sense to a computer. This algorithm includes taking two images
to implement it need to be improved. The accuracy of facial of a person simultaneously and removing the second image from
recognition systems depends on face detection due to this face the first image. This removal causes a discrete boundary outside
detection is the main part of the entire process of the facial the head and if in one of the images the eyes are blinked and
recognition systems there seems to be a little circled region at the eye portions. To
this removed image a connected component procedure is
represented. For the detection of the blinked image, there should
3.1 FINDING FACES VIA COLOR be horizontal and vertical bounded regions. These regions mark
horizontal and vertical segmentation.
A. IMAGES HAVING A DEFINITE BACKGROUD

One process is to find images in which we have a definite

3.3 FINDING FACES IN LIMITED AREAS OF
background containing only grayscale pixels. These images have PIXELS
a narrowband wavelength. While using these images when we
remove the background from the foreground we get facial In the images which are abandoned it is difficult to detect the
boundaries. This is the easiest method for face detection. faces but various methods have now been introduced such as
edge detection orientation, weak classifier cascades. Edge
orientation matching is a technique treated as a template
B. IMAGES HAVING A COLORED BACKGROUND matching procedure. This includes object modeling based on
edge orientations. Various templates are created and matched to
For colored image face detection is based on two procedure the image to detect the edges of the face. It takes approximately
less than 0.08 seconds.
B.1. BY APPLYING A SKIN FILTER

A skin filter is applied to detect skin. The texture of the image 3.4 FACE DETECTION USING HAAR
part that is being masked is defined by the skin filter process. CASCADE
The output generates contains distinct areas of human skin.
Dilation and erosion are the techniques used to develop this kind Haar cascade makes use of the image subtraction philological
of filter. process for face detection. In this, the cascades of distinct
images of the same person are taken and recorded in the
B.2. BY HAULING OUT THE FEATURES WHICH ARE database. All the pixels in the effect of the white region are
BEING MASKED removed from all the pixels in the effect of the black region.
This technique of subtraction is performed on every image in the
During this step, very dark and very bright portions are removed cascade but all the images might not give us the optimal results.
from the image. By removing these portions, we get the most Many of the images have a lot of errors. The image with the
appropriate area covered by the skin for face detection. The minimum error is selected. The result of all the images is added
major problem in this step is the light sources at the time of together and is referred to as a weak classifier. All the weak
facial detection from the camera. Always the camera doesn't classifiers are added together to form a strong classifier.
need to be placed under the sunlight some might have been Applying the subtraction process and determining each image
placed in the average light or low light area. error is a very time and space-consuming process. Instead of
To solve this problem, the input should be in RGB format with applying it on each of the images, subtraction is applied to
good intensity in the range of 0 to 255. images one by one. If the last image is not useful it is discarded.
Haar Cascade is an algorithm for face detection where many
IJCRT2107697 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org g447
www.ijcrt.org © 2021 IJCRT | Volume 9, Issue 7 July 2021 | ISSN: 2320-2882
positive and negative images are used to train the classifier. also be used to detect dead or unconscious individuals at crime
Positive images are those images which we want to identify. scenes.
Negative Images are those images which contains useless things. Gen with 4GB of memory running Windows 10 Home. We used
Pycharm IDE and Python 3.9.1 and OpenCV 4.5.1.48 installed
on my system. The Haar Cascade data file is already provided by
the OpenCV library.

6. EXPERIMENTAL SETUP

6.1 REQUIREMENT

1) Any operating system that will support OpenCV and Python

(Windows, Linux, MacOS)
2) Python
3) OpenCV-Python
4) Haar Cascades Data File
5) i3 or higher core processor (CPU)/ 2.1 GHz or higher
6) Photo/images for testing
Fig 2. Haar Cascades Feature
We used an HP Laptop (15-bs1xx) with a CORE i5 Intel
processor 1.8 GHz of 8th Gen with 4GB of memory running
4. OPENCV STRUCTURE AND CONTENT Windows 10 Home. We used Pycharm IDE and Python 3.9.1
and OpenCV 4.5.1.48 installed on my system. The Haar Cascade
data file is already provided by the OpenCV library.
In our project, we applied face detection to some photos using
OpenCV with Python. OpenCV is an open-source software
library used for computer vision applications. The version we
used of OpenCV for Python called OpenCV-Python because we
developed our project in Python.
We implemented a system for detecting faces in digital images.
These are in JPEG format only. Face detection uses classifiers
algorithms that detect the faces in an image. It has been trained
to detect the face using many images for more accuracy.
OpenCV uses two classifiers, LBP (Local Binary Pattern) and
Haar Cascade classifiers. We use the Haar Cascade classifier.
We already discussed it above.
Fig 3. Structure and content of OpenCV

A. CV- This portion includes image processing and vision

algorithm.

B. MLL- This portion includes statistical classifier and

6.2 FEATURE EXTRACTION-
Process of feature extraction in the Haar Cascade algorithm
clustering tools.

C. High GUI- This portion includes GUI, image and video I/O.

D. CXCORE- This portion includes basic structures and

algorithm, XML support and drawing functions

5. USE-CASES

A. PREVENT RETAIL CRIME- Face recognition is

currently being used to instantly identify when a known thief,
retail criminals, or people with a history of fraud enter retail
stores. According to our data, face recognition reduces external
shrinkage by 34% and, more importantly, reduces brutal events
in retail stores by up to 91%.

B. FIND MISSING PERSONS- Face recognition system is

used to find missing persons. In fact, once in India,
approximately 3000 children were discovered in just four days
with the help of a facial recognition system.
Fig 4. Feature extraction
C. FORENSIC INVESTIGATIONS - Facial recognition
The training data we used is an XML file called:
can help in forensic investigations by recognizing persons in
haarcascade_frontalface_default.xml
security images or other videos. Face recognition software can

IJCRT2107697 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org g448

www.ijcrt.org © 2021 IJCRT | Volume 9, Issue 7 July 2021 | ISSN: 2320-2882
6.3 RUNNING OPENCV 6.5 IMPLEMENTATION FLOW
We prepared a directory where we stored all the files needed.
You will need to put in this directory the following:
1) facial_recognition.py (the name we gave to our program that
contains code. This name can be changed.)
2) haarcascade_frontalface_default.xml (Haar Cascade training
data)
3) Images

WE ARE GOING TO USE THE

1. detectMultiscale: Module from OpenCV to create a

rectangle with coordinates (x,y,w,h) around the face detected in
the image.
2. scaleFactor: The value shows how much the image size is
reduced at each image scale. A small value uses a smaller step
for downscaling. This allows the algorithm to find the face.
3. minNeighbors: It specifies how many “neighbors” each
applicant rectangle should have. A larger value results in small
detections but it detects higher quality in an image.
4. minSize: The minimum image size. By default, it is (30,30).
A Smaller face in the image is best to adjust the minSize value
lower.

6.4 IMPLEMENTATION DETAILS

Facial recognition has three modules. Each module has been
developed by a collaborative effort of all members.

1) Detector: It is responsible to detect faces from the images

captured by a webcam.
2) Data Set Creator: It is responsible to create the data set of
images.

3) Trainer: It is developed to train the software to detect faces.

Fig 5. Implementation Flow

IJCRT2107697 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org g449

www.ijcrt.org © 2021 IJCRT | Volume 9, Issue 7 July 2021 | ISSN: 2320-2882
7. EXPERIMENTAL RESULT 9. REFERENCES
7.1) If the captured image features matched with the image [1] Learning OpenCV –Computer Vision with the OpenCV
stored in the database then it shows ‘Face Match’ Library O’Reilly Publication.

[2] Learning OpenCV: Computer Vision with OpenCV Library,

Kindle Edition. Gary Bradsk1 and Andrian Kehlar

[3] M.A. Turk and A.P. Pentland, “Face Recognition Using

Eigenfaces”, IEEE Conf. on Computer Vision and Pattern
Recognition, pp. 586-591, 1991.

[4] “KyungnamKim” Face Recognition using Principle

Component Analysis”

[5] Codacus: https://youtu.be/1Jz24sVsLE4

[6] G B Huang, H Lee, E L. Miller, “Learning hierarchical
representation for Face verification with convolution deep belief
networks[C]”,Proceedings of International Conference on
Computer Vision and Pattern Recognition,pp.223-226,2012.
7.2) If the captured image features are not matched with the
image stored in the database then it shows ‘Not Face Match’ [7] Computer Vision Papers, http://www.cvpapers.com

[8] Learning OpenCV: Computer Vision with the OpenCV

Library 1st Edition, Kindle Edition

[9] OpenCV Homepage http://opencv.willowgarage.com

[10] Recognition Homepage http://www.face-rec.org/algorithms.

[11] Paul Viola, Matthew Jones Conference paper- IEEE

Computer Society Conference on Computer Vision and Pattern
Recognition.“Rapid object detection using a boosted cascade of
simple features.”

8. CONCLUSIONS
This paper in detail explains the development of a facial
recognition system using OpenCV. We discussed the advantages
of using the OpenCV library in computer vision. The process of
facial recognition with the Haar Cascade algorithm can detect
and recognize the face. The facial recognition process with the
Haar Cascade and can be successfully performed at a distance of
more than 200 cm using a webcam. For future work, while the
current system is for straight faces, it can be improved to
recognize faces at different angles. Followed by the optimization
of the facial recognition process for use on a small mobile
device.

IJCRT2107697 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org g450