You are on page 1of 4

Proceedings of the Third International Conference on Computing Methodologies and Communication (ICCMC 2019)

IEEE Xplore Part Number: CFP19K25-ART; ISBN: 978-1-5386-7808-4

AUTOMATED CRIMINAL IDENTIF ICATION BY FACE


RECOGNITION USING OPEN COMPUTER VISION
CLASSIFIERS
Apoorva.P1 ,Impana.H.C1 ,Siri.S.L1 , Varshitha.M.R1

Ramesh.B2
1
Computer Science and engineering department, MCE, Hassan
2
Professor, Computer Science and engineering department, MCE, Hassan

Abstract-This paper presents a real time Keywords: Automated surveillance


face recognition using a automated camera, face detection, face recognition,
surveillance camera. The proposed Haar cascade, OpenCV.
system consists of 4 steps, including
I. INTRODUCTION
(1)training of real time images (2)face
detection using Haar-classifier (3) Face recognition which is a combination of
comparison of trained real time images machine learning and the biometic
with images from the surveillance techniques which holds the qualities of not
camera (4)result based on the only high precision but also the reliability.
comparison. An important application of For automatically detecting the human’s
interest is automated surveillance, where face from the databases this system can be
the objective is to recognize people who used. In recent years open computer vision
are on a watch list. The aspiration of this has been widely used in different kinds of
paper is to compare an image with applications such as surveillance camera,
several images which has been already robotics etc. This technology is used for
trained. In this paper, we represent a authentication, validation, authorization,
methodology for face detection robustly and identification. In developed countries,
in real time environment. Haar the government creates a datasets which is
cascading is one of the algorithm for face helpful for recognize the human face
detection. Here we use Haar like which compares the suspicious act with
classifiers to track faces on OpenCV trained dataset and information stored in
platform. The accuracy of the face database.
recognition is very high. The proposed Face identification[1] is defined in three
system can successfully recognize more steps (1)face detection (2) feature
than one face which is useful for quickly extraction (3)face recognition. Camera
searching suspected persons as the configuration is very important to track
computation time is very low. In India, moving persons and recognize[2] them
we have a system for recognizing citizen precisely. Facial feature points encode
critical information about face shape.
called Aadhaar. If we use this as a
Precise location and facial feature points
citizenship database we can differentiate
tracing are important. Each feature point is
between citizen and foreigner and
usually detected and traced by performing a
further investigate whether the identified local search for the better matching
person is criminal or not. position[3]. There are very less researches
on face recognition using edge-based

978-1-5386-7808-4/19/$31.00 ©2019 IEEE 775


Proceedings of the Third International Conference on Computing Methodologies and Communication (ICCMC 2019)
IEEE Xplore Part Number: CFP19K25-ART; ISBN: 978-1-5386-7808-4

detection[4]. The edges are not only


carrying valuable data about face but are
also simple to process. The Viola Jones
method builds a classifier by selecting a few
significant features using AdaBoost. Viola
jones method successfully merges more
composite classifiers in cascade
structure[5] which exponentially increases
speed of detector by focusing on the
favourable features of the face.
II. METHODOLOGY
In this paper, we will use three databases.
The images in all the three databases are
preprocessed to remove noise and
redundancy. Then it undergoes feature
extraction where Haar cascade is used. First
database will have the images and details of
all the citizens of the given country. Second
will have the images and details of
criminals/suspects who belong to that
country. Third database will contain the
images and details of the criminals and
suspects who are not the citizens of the
country. The video obtained from the
surveillance camera will be converted into
Fig.1 Flowchart of proposed methodology
frames. When a face is detected in a frame,
it is preprocessed where noise and 2.1 Haar Classifier
redundancies are reduced. Then it
One of the most common tool for face
undergoes feature extraction where Haar
identification is consider as Haar –
cascade comes into picture. The processed
Classifier. Here, the Open Computer
real time image is compared with the
Vision library provides the class for
processed images already stored in the
realize the Haar-Classifier System.
citizen database. If match is found, it is
Though, the most prominent problem in
further compared with the images stored in
the face recognition is the posture of the
local watch list database to identify if the
tilted-face. Here, this manuscript deals
person is criminal/suspect. If he is
with two different resolutions to surpass
criminal/suspect the time for which he was
the problem.
under the surveillance of the camera is
noted. If he is not a citizen, it is further
compared with the images stored in
international watch list database. If match
found, the time for which he was under the
surveillance of the camera is noted. If
match is not found in both the watch lists,
he is considered as an innocent.
Fig.2 Flow Diagram of face detection

978-1-5386-7808-4/19/$31.00 ©2019 IEEE 776


Proceedings of the Third International Conference on Computing Methodologies and Communication (ICCMC 2019)
IEEE Xplore Part Number: CFP19K25-ART; ISBN: 978-1-5386-7808-4

Here, the solutions are described the Python The second one is by creating different
language which contains the syntax that XML files based on the original one. The
shows ease and simple. The 1st solution is new XML files are then used by several
considered as using the more number of new classifiers to recognize object in a
classifiers at the time. Everything should be single image.
trained for detecting and recognizing the
face with various rotation angles. Next the
III. PERIMENTAL RESULTS
2nd solution uses the single classifier for
recognizing many kinds of face images This section pays special attention to the
which has been rotated by various kinds of main consequences of face recognition
angles. Rather than using complex forcriminal identification. This deals with
calculation, Haar-Classifier uses several
outline development and testing the Face
weak classifiers combined to produce the
recognition.
decision. Haar Classifier is also
implemented as a part of Open-CV library, a
popular open source computer-vision library Here, softwares out there that resemble
written in C. the same services as the face recognition
system, however, this also has its own
Open-CV also provides several XML files. merits that can contribute to the society.
Each of them is used to recognize different
Due to its intricacy and revelation
objects such as face, eyes, and mouth.
building, the recognition of the automated
Although the XML files are pretty
face identification tool might br
convenient, they don't cover rotated-object
burdensome. Moreover, there is also
recognition. To recognize rotated-object,
one need to train their own Haar-Classifier possibility for giving the interface as user
to produce suitable XML file. The process – friendly which wants the minimal
is a bit complicated and takes a long time. amount of interaction among the users.
With this system, which automatically
processes the image, users need only
requires to offer input a real time image &
the system will do the rest. Furthermore,
new functions can be added by
experienced developer because this is an
open source system. This is a simple
Fig.3 Haar like features model which always makes the system
easier for utilize.
In this paper, we evaluate two different
As completely, there are some of the
approach to accomplish rotated-object
advantages which have to be recognized
recognition. The first one is by rotating the
image into several different angle, and let as below:
the original Haar classifier do the
recognition task by using the original XML
file.

978-1-5386-7808-4/19/$31.00 ©2019 IEEE 777


Proceedings of the Third International Conference on Computing Methodologies and Communication (ICCMC 2019)
IEEE Xplore Part Number: CFP19K25-ART; ISBN: 978-1-5386-7808-4

[2] Unsang Park, Hyun-Cheol Choi,


It is a healthier substitute which used for Anil.K Jain and Seong-Whan Lee,
identifying the thumb print. Criminal photo “Face tracking and recognition at a
captured through a video source is fed the distance: A coaxial and concentric
identification of the system. This system PTZ camera system”, IEEE
has the capacity for automatically doing the transactions on information forensics
process like recognize, detect and extract and security, vol.8, no.10, October
the features of the images and identifies the 2013
actions. [3] Yongqiang Li,ShangFei Wang,
Yongping Zhao and Qiang Ji,
CONCLUSION AND FUTURE “Simultaneous facial feature tracking
ENHANCEMENT and facial expression recognition”,
IEEE Transactions on image
The proposed face recognition system
processing, vol.22, no.7, July 2013.
based on Haar cascade will be
implemented. Even though the stored set of [4] Reza Moradi Rad,Abdolrahaman
Attar,Reza Ebrahimi atani, “A Robust
images of the person in the database differ
Face Recognition method using edge
from the input image, the system serves as
based features”, 2012 IEEE
a fruitful method of identifying the faces.
Symposium on Computers and
Inorder to find the difference between input
Informatics.
image and stored images the system takes
[5] Paul Viola, Michael Jones, “rapid
out and calculates main features from the
Object Detection using a Boosted
input image.
Cascade of Simple Features”.
Thus, some changes in the new face image
to be recognized can be allowed.
Recognition precision and better
discriminatory power, computational cost
because smaller images with only main
features require least processing to train the
Haar wavelets. The main advantage is that
we use citizenship database which already
exists.

REFERENCES

[1] Jyoti Nautiyal, Shivali Gahlot and


Pawan Kumar Mishra, “An
automated technique for criminal
face identification using biometric
approach”,Conference on
Advances in Communication and
Control Systems 2013(CAC2S
2013)

978-1-5386-7808-4/19/$31.00 ©2019 IEEE 778

You might also like