COMP8701: Multiple Face Recognition Systems Research Paper

This research paper delves into the realm of multiple face recognition systems, exploring their significance in modern security applications. The paper begins with an abstract and introduction, highlighting the technology's importance in biometric systems and its advantages over alternatives like fingerprint and signature scanning. It discusses the core components of face detection and recognition, and examines the historical development of the technology, including early methods and the advancements in 3D face recognition. The paper also provides an in-depth analysis of the algorithms used in face recognition, including the steps involved in face detection (filtering, resizing, color mode conversion, morphological operations, and non-face region elimination) and face recognition (lightning effect, scaling, correlation, and segmentation). The report reviews relevant literature and concludes by emphasizing the technology's increasing adoption in various organizations and event management security, owing to its superior performance compared to other security systems. The paper includes references to published papers and adheres to the guidelines provided for a conference-style research paper.

Running head: MULTIPLE FACE RECOGNITION SYSTEMS
MULTIPLE FACE RECOGNITION SYSTEMS
Name of the Student
Name of the University
Author Note

1MULTIPLE FACE RECOGNITION SYSTEMS
Abstract
At present security issue is one of the major concerns in organizations and being the top
priority firms’ are ready to invest maximum on their security techniques and for this purpose
new technologies are being adopted and trained personnel hired to achieve the security that is
highly desired. Technologies are replacing their human counterparts rapidly due to their
perfection and reliability. Multi face recognition system is a revolutionary innovation related
to information processes based on biometrics and is considered to be more effective than its
other contemporaries like fingerprint scanning and signature scanning. The technology makes
use of two vital techniques that are face detection and face recognition. In this paper the topic
of discussion is the process of development of this modern technology and its applications in
the security domains. The history and the methodology of the multi face recognition system
will also be discussed. Lastly the paper concludes by commenting on the future impacts of
the multi face recognition system and the present scenario based on the discussions made.

2MULTIPLE FACE RECOGNITION SYSTEMS
Introduction
Multi face recognition serves as a major element in the sphere of information
biometric systems that deals in detecting and recognizing individuals according to their
unique character traits. This particular technology helps recognize an individual by his facial
characteristics and is a rapidly growing technology (Wen et al. 2016). The security
implications of the technology will be one of the main subjects of the following paper along
with the detailed methodology and the algorithms used to achieve the desired results.
Literature review based on the evolving history and the software involved will also be
considered in this paper. The technology has a vital impact on society owing to its security
implications. Multi face recognition has been considered effective due to its broad range of
applications and efficient working process. Security being the topmost priority in every
business and service organizations the above technology has been adopted at a faster rate as
no one wants to compromise with issues related to the security of their firm (Aithal 2015).
Biometric technologies are finding their applications in many domains especially in the
security domain and face recognition are one of the major advancements in this arena. The
efficiency of the technology to recognize the faces of the individuals makes it globally
important as it serves as a key remedy for the prevalent security issues. The key terms in the
topic are digital image processing, biometrics, face detection and face recognition. The
concept used in achieving the desired solutions to the security framework of an event security
management based on this technology is to be discussed.
Discussion
There have been much software developed in context of face recognition during the
last decades and different methods are used by each of the software based on the unique

3MULTIPLE FACE RECOGNITION SYSTEMS
algorithms inculcated in them. For example some of the facial recognition software works on
the methodology of extracting the traits of the face of an individual in order to indentify the
individuals face and other algorithm normalize a definite set of images compressing the data
and later saving the same in a single image that is used to recognize the face of an individual.
This image that was compressed earlier to a single one is compared with the data recorded
related to face thus recognizing the person concerned. The history of the face recognition
technology dates back to as old as 1964-1965 when it was being researched on by an
American mathematician Woody Bledsoe (Bhati and Gupta 2015). Initially the recognition
was based on extracting the landmarks of the face portions such as mouth, area near and
around the eye. The database recorded is huge and it was difficult to sort the images thus
creating a problem and forcing the scientists to search for new developments in the existing
technology. With the advancements in this field a modern method that is being used is the 3D
face recognition system. It utilizes a 3D sensor to record or capture information related to the
shape of the face and in this case the unique and distinct feature of the facial structure is used
for the detection and recognition (Lohiya and Shah 2015). The developments in this specific
technology have some advantages over the traditional ones mainly that the algorithms used
do not get affected changing light intensity and detection and recognition of the features can
be done based on the wide range of angles that includes the view of the profile.
Algorithms of the face recognition system
The algorithms of the multi face recognition system consist of two main parts and
they are face detection and normalization and face identification. The algorithms consisting
of both these parts are considered to be fully automatic and those consisting of the face
identification part only are termed as partially automatic algorithms. The difference between
the partially automatic algorithm and the fully automatic algorithm is that the former uses
facial image along with the specific coordinates of the eyes center while the latter makes use

4MULTIPLE FACE RECOGNITION SYSTEMS
of the facial images only (Mlakar and Potočnik 2015). The rapid advances in this domain
have provided the organizations to choose from three different types of algorithms related to
recognition that include the profile, view-tolerant and frontal recognition techniques which
depend on the images type and the various recognition algorithms. Frontal recognition is the
classical way of approach while view-tolerant algorithm is a sophisticated way that considers
some specific underlying geometry, statistics and physics. The profile schemes have marginal
significance in case of identification. The systems follow the algorithms for both the
detection and then recognition. The algorithms related to both the operations are discussed
below.
. First considering the face detection case, the necessary steps are filtering, resizing,
color mode conversion and skin detection, morphological operations and lastly elimination of
the non face regions. The steps in details are as follows:
a) Filtering: The process includes taking the first image and then the technique of median
filter is applied to it. The aim of the technique is eliminating the noise that appears during
capturing of the image and at the same time enhancing the procedure of edging.
b) Resizing: After the image is filtered the next process involves resizing from the resolution
of the camera t0 512*384*3 as it is easier to perform MATLAB operations in order to deal
with the smaller images and in this way the processor will also work faster.
c) Color Mode Conversion and Skin Detection: The image that had been taken initially
from the lenses follows the RGB color code that is bad to analyze, as it mixes the luminous
and the cruminous and thus makes the removal of the important parameters difficult, so the
picture requires to be converted to HSV that facilitates process involved in considering the
essential parameters (Patil, Kothari and Bhurchandi 2015). HSV here stands for hue,
saturation and value and also called HSB that stands for hue, saturation and brightness.

5MULTIPLE FACE RECOGNITION SYSTEMS
d) Morphological Operations: This operation includes three basic steps namely dilation,
filling and erosion (Batool and Chellappa 2015). The three steps are as follows:
1. Dilation- Dilation refers to the process involved in conversion of black color having value
0 to the white color that has value 1. It applies mainly to binary images but there are versions
available that can also work on the grayscale images. The primary effect of the operation is to
enlarge the perimeters of the areas of the foreground pixels. The regions of the foreground
pixel enlarge in size the holes within the areas become smaller. Two inputs in this case is two
pieces of the data that consist the image needed to be dilated and the other is a set of points or
coordinate points that are known as structuring element.
2. Filling- The process involves filling a hole near the vicinity of a white color with white
color having value 1. It can be referred to as transformation of black pixels into white ones
that exist near the white area. This case has a link with the algorithm of skin detection as the
eyes could not be considered as skin pigments thus they were converted into black ones. In
the MATLAB operations this filling function takes an input of a binary image and the images
with the filled holes are the outcomes.
3. Erosion- It is used as a dilation operator, typically applies to the binary images. It erodes
away the boundary of foreground pixels mainly the white pixels thus shrinking the areas
around the foreground pixel, and enlarging the holes within that region (Azad, Ahmadzadeh
and Azad 2015). It is similar to the dilation operator and takes two inputs as in case of
dilation.
e) Elimination of the non-face regions- Two techniques are involved to perform the
elimination of non-face regions that include the pixels and the ratio technique. A genuine face
has 500 pixels and when a skin area less than 500 pixels is detected it gets eliminated
automatically. In the ratio technique a face must have a height-width ratio in between 0.4 to

6MULTIPLE FACE RECOGNITION SYSTEMS
2.5 and in case of any discrepancy in the said value it is eliminated. After the techniques are
performed the face area remains and the unwanted part eliminated.
The face recognition algorithm consists of lightning effect, scaling, correlation and
segmentation. In lightning effect the light effect in the image will be reduced to match with
the image stored in the database. Scaling refers to scale the image captured to same size as
that of the image stored in the database. Correlation is done after scaling and if the value of
the correlation is in between the range 0.5 to 0.9 the process halts otherwise segmentation is
opted for that divides the face into three regions and then each segment is correlated with the
captured image (Haghighat, Abdel-Mottaleb and Alhalabi 2016).
Conclusion
On the basis of the discussion it can be concluded that face recognition systems are
finding their applications in the society and mainly in the security aspects owing to the better
performance they are providing when compared to other security systems. The system
involves complex algorithm structure that increases its efficiency and the developments in the
technology are making it popular in this domain resulting it their adoptability in various
organizations and event management security.

7MULTIPLE FACE RECOGNITION SYSTEMS
References
Wen, Y., Zhang, K., Li, Z. and Qiao, Y., 2016, October. A discriminative feature learning
approach for deep face recognition. In European conference on computer vision (pp. 499-
515). Springer, Cham.
Aithal, P.S., 2015. Biometric Authenticated Security Solution to Online Financial
Transactions. International Journal of Management, IT and Engineering, 5(7), pp.455-464.
Lohiya, R. and Shah, P., 2015. Face recognition techniques: A survey for forensic
applications. International Journal of Advanced Research in Computer Engineering &
Technology (IJARCET), 4(4).
Bhati, D. and Gupta, V., 2015. Survey-A Comparative Analysis of Face Recognition
Technique. Int. J. Eng. Res. General Sci., 3(2), pp.597-609.
Mlakar, U. and Potočnik, B., 2015. Automated facial expression recognition based on
histograms of oriented gradient feature vector differences. Signal, Image and Video
Processing, 9(1), pp.245-253.
Patil, H., Kothari, A. and Bhurchandi, K., 2015. 3-D face recognition: features, databases,
algorithms and challenges. Artificial Intelligence Review, 44(3), pp.393-441.
Batool, N. and Chellappa, R., 2015. Fast detection of facial wrinkles based on Gabor features
using image morphology and geometric constraints. Pattern Recognition, 48(3), pp.642-658.
Haghighat, M., Abdel-Mottaleb, M. and Alhalabi, W., 2016. Discriminant correlation
analysis: Real-time feature level fusion for multimodal biometric recognition. IEEE
Transactions on Information Forensics and Security, 11(9), pp.1984-1996.
Azad, R., Ahmadzadeh, E. and Azad, B., 2015. Real-time human face detection in noisy
images based on skin color fusion model and eye detection. In Intelligent Computing,
Communication and Devices (pp. 435-447). Springer, New Delhi.

8MULTIPLE FACE RECOGNITION SYSTEMS