ISSN 2321 3361 © 2021 IJESC

Research Article Volume 11 Issue No.05

A General Review of Human Face Image Detection using Machine

Learning Classifier
Shubhangi B. Kamble1, Dr. V. D. Jadhav2
P. E. S College of Engineering Aurangabad, Maharashtra, India

Abstract:
Face image detection is widely used technology has proven for many application like security camera, checking system at
airports, retail marketing, people counting, mall etc. Face image detection is a very important field now a days because all over
world used different face detection software application .This technology is only finding the human faces image in a frame
video and counting the images. In this research, an efforts to use the haar cascades classifier which is comes from viola jones
algorithm haar cascades method used the haar like feature by using this feature is to finding the efficient human feature like
eye, nose, and forehead etc. In this study we uses the viola-jones algorithm for face detection. This system proposed two
methodology to find the face, first phase is to detect face in clustered image using viola-jones. Second phase this algorithm is
to extract the feature from face and classifies these feature using add boost classifier. This system, aim to determine the
existence of the position of the human face in a frame. This study helps to person’s identification/identity in various
application like home security, camera surveillance in any field.. In this research discovering the how works the viola-jones
algorithm, finding the human face accurately with contour and real time face detection. We use this system in many social
media application.

Keywords: Viola-jones, Haar like feature, Ada boost classifier, face detection, Open CV.

I. INTRODUCTION identify the presence of people’s faces within digital images.

In order to work face detection applications uses machine
Object detection is the one of the computer technology learning and is a bordered term than face recognition. Face
which is the connected to the digital image processing and detection just means that a system is able to identify that
computer vision [1]. Face is the one of the outer part of the there is human face in an image and video. Face detection
body that is very important because it can be used as has several application, only one of the face recognition and
person’s identification /identity. Various methods used to also used to auto focus cameras.
facial feature tracking. In this paper we, using the haar
cascade classifier which is the part of the viola-jones II. LITURATURE REVIEW
algorithm. The aim of this algorithm is to finding the
position of human face image in a frame and capable of When now a day’s face detection [6] and face recognition is
detecting the face in image with high accuracy [2]. It is the widely used method in a real world application. This all
connected with the automatic identification of an individual work will be done by using viola jones [3] algorithm has
in a digital image. Face detection [6] is used to identifying made face detection feasible in digital cameras and video
and locating the human face irrespective of its size, position stream, image organization software. In this research, we
and situation in a frame and also camera. Detection of human presented real time human face detection like social media
face is the very easy task for human brain but very, it is very application based on viola-jones approach. The rest of this
difficult task for computer system [3]. From last few years, article organized by follow. Another section we will give the
the face detection is very important at this era, because it can brief overviews of the feature selection for our face detection
be implemented in various field such as photography, task, especially haar-like feature and its computational
security system and databases. To detect the face is easily framework. We focus on Adaboost learning [4] classifier to
and accurately computer system needs some training factors train a strong classifier for face detection. They [Komukai
so that the computer system can be easily identify the face or Toshiba-Cho] et .al proposed in that paper generic and
non-face. Detecting the face we need some thresholds value efficient objection frame work and tackle the cat face
these value have sets, based on these values a system can detection by using haar-like feature. [1] There have been
detect human face. If the image specifies the desired proposals to use vehicle detection technique that can be used
threshold value then the image is face otherwise it is a non- for traffic surveillance system equipped with electronic
face. Face detection[6] divided in two parts:-1. Features devices traffic condition, condition monitor rules and
based:-depending on some simple features like face variation regulation and avoid collision between cars. [Shaif
and posture. 2. Learning based: - use an amount of training Choudhury] et.al [2] in this research represented to the face
models and machine learning algorithm. [2]. Face detection identification this technology based on the identification
technology refers to computer technology that is able to card.by using Adaboost learning [4] algorithm utilizes

IJESC, May 2021 27963 http:// ijesc.org/

simple features to build cascade classifier for face detection 4. RGB Image
having more advantage like speed for identification [Wang RGB Image are represented as a combining the 3 colors Red,
Chongwen] et.al [3] in general, the emotional state is defined Green, and Blue. RGB image consist of 8 bit per pixel. In
as 7 like that neutral, happy, sad, angry, fearful, surprised, RGB Image have different shades of colors can represented
disgusted. FER [facial expression] can be used analyze the with 0 consist of black and 255 white. Fig [2]
human face expression. These human FER used in such as
in driving situation especially motion slowdown due to the
related with traffic rules. [Sanghyuk Kim] et.al [4] adult
image detection using semantic matching is a hard task to
defining the online skin tone sampling mechanism based on
face detection. The skin tone sampling chromatic distribution
from the face and the skin chromatic distribution is
employed to filter out the non-skin image of target image.
[Jiann-shu lee] et.al [5]. Abdu Salim mullah and Asif Anjum
Akash et.al[6] proposed in this research of human face
detection [6] from digital image or video stream is used but
sometimes a system detect an object where there is no face at
all they improvement to haar features to reduce such wrong
detection rate introducing human skin color (HCS)
characteristics. Earlier, some research studies based on face
detection using haar cascade classifier which is part of viola- Figure. 2.RGB model
jones algorithm; however past literature doesn’t provide to
evidence of real time human face detection by using 5. Grayscale Image: -
Aadboost classifier. Grayscale Image also consist of 8 bit per pixels. This
Grayscale image have 256 different shades like that RGB 0
III. METHODOLOGY& EXPERIMENTATION pixel define black and 255 denotes white.

Basic Terminology

1. Face
Face is the central organ of the human, to expression
recognition and human communication. Face have four
sensory organs. There are nose, ears and tongue. Face is the
Front part of the head and extends from the forehead to the
chin includes mouth, cheeks, and eyes.

2.Computer vision
It is the field of computer science that include methods to
obtaining, managing, and analyzing visual data such as image
and video. This field related to the computer vision are image Figure. 4.Gray Scale Image
processing and machine vision. Fig [1]
6. Viola-Jones Algorithm
Viola-jones [3] algorithm is a very popular algorithm for
face detection[6]. It is the object detection framework this is
used in real time face detection and training rate is high
when four major ideas come in Viola-Jones face detection
method. 1. Haar like features 2.Integral image 3.Adaboost
machine learning [4] and 4. Cascade classifier.

6.1 Haar features: -

Haar features are digital image features used in any object
detection. All human faces have some universal properties
like eyes region is darker than its neighbor pixels; and the
nose region is brighter than the eye region easy way to find
out which region is lighter than brighter is to sum up the
Figure.1. Computer vision pixel values of both region and compare them. Elaborate the
sum of pixels values in the darker region will be smaller than
3. Open CV Library the sum of pixels in the lighter region. If one region is lighter
(Open source computer vision Library)[5]Is an open source than the other, it may be an edge of an eyebrow or
computer vision and machine learning software library called sometimes middle portion may be shinier than surrounding
Open CV [5]. This library developed by Intel Corporation boxes which can be interpreted that time we uses Haar like
which is specially used in image processing features. There are 3 types of haar like feature.

IJESC, May 2021 27964 http:// ijesc.org/

 In Viola-jones [3] algorithm each Haar like feature
represents a week learner to decide the type and size of
feature that goin to the final classifier. Adaboost [4] cheeks
to the performance of all classifier that you supply to it after
that section.

4. Cascade classifier
AdaBoost [4] will finally selected the best features around
say 2500, but it is still a time consuming process to calculate
these features for each region. In that window detector which
we slide over the input image and they find only those region
which is face contain. The job of the cascading is to quickly
discard the non-face, and avoid wasting precious time and
computation. This is an achieving the speed necessary for
real – time face detection. Fig [3]
Calculating the haar features in single number the sum of
pixel value in the black area minus the sum of pixel value in
white area.

6.2 Integral image

In that integral image we use value of each point is the sum
of all pixels above and to the left including the target pixel
Figure.5. Cascade classifier
0 1 1 1
1 2 2 3 IV. CONCLUSION
1 2 1 1
This system developed a fused face detection and
1 3 1 0
recognizing system by using machine learning algorithm.
Improve speed of human image recognizing. Face detection
6.2.1 Original image system are going to have widespread application in smart
Using to integral image we save a lots of time calculating the environment. Face recognition technology has come a long
summation of all the pixels in rectangle as we only way in the last twenty years. Today, machines are able to
calculating on four edge of the rectangle. automatically verify identify information for secure
transactions, for surveillance and security tasks, and for
0 1 2 3 access control to buildings etc. These application usually
1 4 7 11 work in controlled environments and recognition algorithms
can take advantage of the environmental constraints to obtain
2 7 11 16 hung recognition accuracy. The next generation face
3 11 16 2 recognition system are going to have widespread application
in smart environments. To active this goal computer must be
able to reliably identify nearby people in a manner that fits
6.2.2 Integral image naturally within the pattern of normal human iterations. This
When we add the pixel in blue box we get 8 as the sum of all system must not require special interactions and must
pixels 2+1+1+3+1+0=8 we had to use six elements involved conform to human intuitions about when recognition is
in your calculation now we calculate the sum of these same likely.
pixels using the integral image you just need to find the
corners of the rectangle and then add the vertices which are ACKNOWLEDGEMENT
green and subtract the vertices in the red boxes, now doing
that 21+1-11-3=8. We get the same answer and only four We want to thankful for department of computer sciences
number are involved in calculation. No matter how many and engineering of P.E.S College of engineering aurangabad
pixels in that rectangle box. We just need to find corners of for the support and guidance. Special thanks to my guide and
rectangle.Then this technique apply in to calculating the value my family for their belief and support.
of the Haar features. You have simple way to calculate the
difference between the sum of pixels black region and white V. REFERENCES
region.
[1]. 10 January 2021) D. Dhivehi, "Towards Data Science,”
3. AdaBoost learning [online]. Available
AdaBoost learning algorithm. In 24*24 window detector is https://www.google.com/search?q=towards+data+science&rl
nearly 160,000 features, but only a few of these features will z=1C1CHBD_enIN942IN942&oq=Towards+Data+Science&
be important to identify the human face. So that’s why used aqs=chrome.0.0i433j0l9.1314j0j15&sourceid=chrome&ie=U
Adaboost learning [4] algorithm to identify the best feature TF-8l
in 160,000 features.

IJESC, May 2021 27965 http:// ijesc.org/

[2]. (February 2020). “APA itrajah: Anatomy, Fungi, and
DanProcedureTerabit”. Dodo Corporation, [Online].
Available:http://freakonomics.blogs.nytimes.com/2008/09/09
/the-economic https://www.google.com/satment&st=cse

[3]. P. Viola and M. J. Jones. (20 march 2021). “Rapid object

detection using a boosted Cascade of simple features”, IEEE
Conference on Computer Visionand Pattern Recognition.

[4].You Freund and Robert E. Schapiro. (12 January 2021).

“A decision-theoreticGeneralization of on-line learning and
an application to boosting. In Computational Learning
Theory: Euro colt ’95, pages 23–37.Springer-Verlag, 1995
ence.”

[5]. A. Murine. (2021, March 12). “] Opens face detection.”

[Online].Available:
https://towardsdatascience.com/face-detection-in-2-minutes-
using-opencv-python-90f89d7c0f81 .html

[6]. (12, 2021, may). A snow based face detector. In Neural

Information Processing.” Roth, M. Yang, and N. Ahuja.
[Online].Available:
https://www.google.com/search?q=A+snow+based+face+det
ector.+In+Neural+Information+Processing.%E2%80%9D&rl
z=1C1CHBD_enIN942IN942&oq=A+snow+based+face+det
ector.+In+Neural+Information+Processing.%E2%80%9D+&
aqs=chrome..69i57.1370j0j15&sourceid=chrome&ie=UTF-8l

[7]. (12 may 2021).” Moving object detection and tracking

for video surveillance.” object detection available [Online],
Available:
http://ijergs.org.managewebsiteportal.com/files/document
s/MOVING-46.pdf 7:20l

[8]. (12 April 2021). “Fast training and selection of Haar

features using statistics in boosting-based face detection
[Online]. Available:
7:50https://ieeexplore.ieee.org/document/4409038 org/x

IJESC, May 2021 27966 http:// ijesc.org/