Post on 10-Jul-2020
transcript
ADVANCES in NATURAL and APPLIED SCIENCES
ISSN: 1995-0772 Published BYAENSI Publication EISSN: 1998-1090 http://www.aensiweb.com/ANAS
2017 May 11(7): pages 139-144 Open Access Journal
ToCite ThisArticle: P. Arun Mozhi Devan, M. Venkateshan, A. Vignesh, S.R.M. Karthikraj., Smart Attendance System Using Face Recognition. Advances in Natural and Applied Sciences. 11(7); Pages: 139-144
Smart Attendance System Using Face Recognition
1P. Arun Mozhi Devan, 2M. Venkateshan, 3A. Vignesh, 4S.R.M. Karthikraj
1Assistant Professor, 2,3,4Student Scholars Electronics and Instrumentation, Sri Ramakrishna Engineering College, Coimbatore. India. Received 28 Feb 2017; Accepted 14 May 2017; Available online 19 May 2017
Address For Correspondence: P. Arun Mozhi Devan, Assistant Professor, Student Scholars Electronics and Instrumentation, Sri Ramakrishna Engineering College, Coimbatore. India. E-mail: arunmozhideven.p@srec.ac.in
Copyright © 2017 by authors and American-Eurasian Network for ScientificInformation (AENSI Publication). This work is licensed under the Creative Commons Attribution International License (CC BY). http://creativecommons.org/licenses/by/4.0/
ABSTRACT Most representative part of body can be used to differentiate one person among others is face and it is a fundamental technology of biometrics, which has been applied to a variety of areas like computer vision, security systems (pattern recognition), human machine interaction and image processing. In this proposed approach the features of the query image (current facial image) and database images (stored facial image) features have been extracted by Viola Jones algorithm and trained using ANN. This paper is to increment the output neurons simultaneously with incrementing the input patterns started with a small number of output neurons and a single hidden-layer using an initial number of neurons. Face recognition experiments were carried out by using Artificial Neural Networks. The main objective of this proposed system is to develop an efficient face recognition system by improving the efficiency of the existing face recognition systems and also for the process of secured attendance system. This attendance system uses the mapping of the images with the database and it updates the days of present and also notifies the external person’s presence.
KEYWORDS: Viola Jones algorithm, vital feature points,AdaBoost classifier, face identification, facial features, mapping,
Artificial Neural Networks (ANN)
INTRODUCTION
A facial recognition system is a computer application capable of identifying a person from a digital image
or a video frame from a video source. One of the ways to do this is by comparing selectedfacial features from
the image and a facial database. An Attendance Management System developed using biometrics generally
consists of Image Acquisition, Database development, Face detection, Pre-processing, Featureextraction and
Classificationstages followed by Postprocessingstage. Traditionally attendances are taken manually by using
attendance note available for class, office, etc. which is a time-consuming event. Moreover, it is very difficult to
verify one by one in a large environment withdistributed branches whether the authenticated persons are actually
responding or not. Other biometrics likefingerprints, iris scans, and speech recognition cannot performthiskind
of mass identification.Inthispaper, proposed system takes the attendance of peopleduringworking hourswhich
has been implemented in MATLAB software.
This system takes theattendance automatically using face recognition. However, it is difficult to estimate
the attendance precisely usingeach result of face recognition independently because the face detection rate is not
sufficiently high.The proposed methodestimates the attendance precisely using all the results of face recognition
obtained bycontinuous observation and it improves the performance for the estimation of the attendance.
Thispaper first review the related works in the field of attendance management and face recognition. Finally,
experiments are implemented to provide as evidence to support our plan. Theresult shows that continuous
observation improved the performance for the estimation of the attendance.
140 P. Arun Mozhi Devan et al., 2017/Advances in Natural and Applied Sciences. 11(7) May 2017, Pages: 139-144
II. Related Work:
In [1] the authors have proposed a finger print based attendance system. A portable fingerprint device has
been developed which can be passed among the students to place their finger on the sensor during the lecture
time without the instructor’s intervention. This system guarantees a fool-proof method for marking the
attendance. The problem with this approach is that passing of the device during the lecture time may distract the
attention of the students.
In [2] the authors have proposed RFID based system in which students carry a RFID tag type ID card and
they need to place that on the card reader to record their attendance. This system may give rise to the problem of
fraudulent access. An unauthorized personmaymake use of authorizedID card and enter into the organization.
In [3] the authors have proposed Dousman’s algorithm based Iris recognition system. This system uses iris
recognition management system that does capturing the image of iris recognition, extraction, storing and
matching. But the difficulty occurs to lay the transmission lines in the places where the topography is bad.
III. Proposed Framework:
The proposed automated attendance management system is based on face recognition algorithm. When a
person enters the room his image is captured by the camera at the entrance. Face region is then extracted and
pre-processed for further processing. Face Recognition proves to be advantageous than other systems are
discussed in the Table I. The stages in the proposed Automated Attendance Management System are as shown
in the Fig.1 Technical details of implementation of each stage are discussed in the next sections.
Fig. 1: System block diagram
Table I: Comparison Of Various Attendance Systems Type of the System Drawback
RFID-based Fraudulent usage
Finger print based Time Consuming for people to wait and give their attendance
Iris-based Invades the privacy of the user
Wireless-based Poor performance if topography is bad
A. Feature Points on Human Face:
Applying human visual property in the recognition offaces, people can identify face from very far distance,
even the details are vague. Human faces made up of eyes, nose, mouth and chin etc. Thereare differences in
shape, size and structure of thoseorgans, so the faces are differ in thousands ways. One commonmethod is to
extract the shape of the eyes, nose, mouth and chin, and then distinguish the faces bydistance and scale of those
organs.Wecan normalize the faces in the database through pre-treatment, so as to extend the range of database,
reduce the storage and recognize the faces moreeffectively.
141 P. Arun Mozhi Devan et al., 2017/Advances in Natural and Applied Sciences. 11(7) May 2017, Pages: 139-144
Additionally, the selection of face feature point’s iscrucial to the face recognition. The number of the
feature points should take enough information and not be too many. If the database has different post rues of
each people to be recognized the property of angle in variance of the geometry characteristicis
veryimportant.This paper haspresented a method to locate the vital feature points offace, which select 9 feature
points [4] that have theproperty of angle invariance, including 2 eyeballs, 4near and far corners of eyes, the
midpoint of nostrilsand 2 mouth corners, as shown in Fig.2
Fig. 2: The 9 vital feature points on face with extracted value of image
Where,
C1 = Left eye center-Eye center
C2 = Right eye center-Eye center
C3 = Left eye center-Mouth center
C4 = Right eye center-Mouth center
C5 = Left eye center-Nose center
C6 = Left eye center-Nose center
B. Face Detection:
A proper and efficient face detection algorithm alwaysenhances the performance of face recognition
systems.Various algorithms are proposed for face detection such asFace geometry based methods, Feature
Invariant methods,Machine learning based methods. Out of all these methodsViola Jones [5] proposed a
framework which gives a highdetection rate and is also fast.Viola-Jones detection algorithm [10] is efficient for
real timeapplication as it is fast and robust. Hence we choseViola-Jones face detection algorithm which makes
useof Integral Image and AdaBoost learning algorithm asclassifier [6]. We observed that this algorithm gives
betterresults in different lighting conditions and we combinedmultiple Haar classifiers to achieve a better
detection ratesup to an angle of 55 degrees.
C. Database Development:
As we chose biometric based system enrolment of everyindividual is required. This database development
phaseconsists of image capture of every individual and extractingthe bio-metric feature, in our case it is face
features and laterit is enhanced using pre-processing techniques and1 storedin the database. In our project we
have taken a database of13individuals with6 images of eachhas been collected for this project.
D. Feature Extraction and Classification:
The performance of a Face Recognition system also depends upon the feature extraction and their
classification to get the accurate results. Feature extraction is achieved using feature based techniques or holistic
techniques. In some holistic techniques we can make use of dimensionality reduction before
classification.Principal Component Analysis (PCA) was the first algorithm that represents the faces
economically. In PCA the face images are represented using Eigen faces and their corresponding projections
along each Eigen face. Insteadof using allthe entire dimensions of an image only meaningful dimensions are
considered to represent the image. Mathematically an image using PCA is represented as
X= WY + μ
WhereXis the face vector, Y is vector of Eigen faces, W isthe feature vector, and μ is the average face
vector. In general features extracted from PCA and Linear Discriminant Analysis (LDA)is subjected to distance
classifiers [7]. The distance between the features of probe image and features of trained images is
calculated.PCA is used for feature extraction and Support VectorMachine (SVM) is used for the
142 P. Arun Mozhi Devan et al., 2017/Advances in Natural and Applied Sciences. 11(7) May 2017, Pages: 139-144
classification.For recognition SVM findsthe optimal separation of closest points in the training set and we
require a multi-class classification. SVMSupport Vector Classification atype is used formulti-class
classification. In classification of imagessmall amount of trainingdata [8] is enough for estimation. So Face
Recognition involves in two stages, featureextraction and classification. The above mentioned featureextractors
combined with classifiers are compared invarious real world scenarios such as lighting conditions,Unintentional
facial feature changes and Expressions.
E. System Accuracy:
System Performance is also evaluated in termsof recognition rate, distance, false positive rate, timetaken for
training. This system can be employed for ‘n’ number of peoples, but for our demonstration we have taken our
class students as an example and False Positive Rates are calculated byconsidering 13 real time image frames
and system accuracy will be given as system accuracy and it will be denoted as
𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 =TP+TN
TP+TN+FP+FN (1)
TP defined that face is correctly identified (out of 40 correct faces for the same person); TN implies that
faces is correctly rejected (out of remaining 25 wrong persons). FP imp lies that face is incorrectly identified,
and FN equals face is incorrectly rejected. Our aim is to maximize TP and TN and minimize FP and FN. This
can be explained with the MSE of NN and the number of epochs taken to converge with the desired goal of our
trained value and it is shown in Fig.3.
Fig. 3: Performance and accuracy of trainedset of images
Training state results of the neural network (in epochs) has been shown in Fig.4.Distance also playsas a
criterion in this system model as the face image frames arecaptured from the video and face region isresized. So
the face region captured at about 4feet to17feet gives better results.For a Training data of 13 images trainingtime
is calculated.Algorithm requires minimumtime for training where as SVM and Bayes classifiers takemore time
for training. In classifiers comparison SVM doesbetter classification than the rest.Dewi’s [9] system accuracy is
around 93.7 % while the proposed system accuracy is 98.14 %. Regarding computation time, their system spent
3.52 seconds while the proposed system takes 1.56 seconds.
Fig. 4: Training state results of the neural network (in epochs)
143 P. Arun Mozhi Devan et al., 2017/Advances in Natural and Applied Sciences. 11(7) May 2017, Pages: 139-144
IV.Results:
In the proposed system, after recognizing the faces ofthe students (shown in Fig.5) the students roll
numbers are updated into an excel sheet shown in Fig.6.At the end of the class a provision to announce
thenames of all students who are present in the class isalso included.
Fig. 5: Recognized faces of the students
The system is also equipped with the facilityof sending notification messages to the absentees when
thatfacility is enabled.For detecting face and its features using the Viola Jones cascaded object detection
framework, it takes 0.8168 seconds toprocess single 640 x 480 pixels image.This system also notifies the
unwanted person’s like lecturers, other class students and workers (Untrained Faces)interference with in the
class which is shown in Fig.7
Fig. 6: Updated attendance into an excel sheet
Fig. 7: Untrained face popup notification
144 P. Arun Mozhi Devan et al., 2017/Advances in Natural and Applied Sciences. 11(7) May 2017, Pages: 139-144
V.Conclusion And Future Scope:
Current work is focused on the method to obtain the different weights of each focused face based on
location. The proposed methods have provided rather good facial feature detection. Examples demonstrating the
success of the various proposed methods as well as criticism when they fail have been included. We proposed a
complete framework for accurate human face-based identification system. Face not only detected using this
algorithm but also the distance of the facial characters. Also it is detected in both colour and grayscale images
under varying conditions.
Success rate at face recognition is around 93% to 95% and regarding face identification is 99%. The system
error rate was around 5-7%. Using 6 face distances was good enough to distinguish faces, and using more
distances can make result better. This face recognition also used for the process of making the attendance of the
students.In future multiple face recognition can be improved by integratingvideo-streaming service and lecture
archivingsystem, to provide more profound applications in thefield ofdistance education, course management
system(CMS) and support for faculty development (FD).
REFERENCES
1. Mohamed, B.K. and C. Raghu, 2012. “Fingerprint attendance system for classroom needs,” in India
Conference (INDICON), pp: 433-438.
2. Lim, T., S. Sim and M. Mansor, 2009. “RFID based attendance system”, Industrial Electronics
&Applications,IEEE Symposium, 2(3): 778-782.
3. Kadry, S. and K.S. maili, 2007. “A design and implementation of a wireless iris recognition attendance
management system”, Information Technology and control, 36(3): 323-329.
4. Noura, A Semary and Ahmed Fawzi Gad, 2014. “A Proposed Framework for Robust Face Identification”,
System Computer Engineering and System (ICCES), pp: 62-67.
5. Viola, P and M.J. Jones, 2004. “Robust Real-Time Face Detection”,
International Journal of Computer Vision, 57(2): 137-154.
6. Amr El Maghraby, Mahmoud Abdalla, Othman Enany Mohamed and Y. El Nahas, 2013. “Hybrid Face
Detection System Using Combination of Viola - Jones Method and Skin Detection”, International Journal
of Computer Applications, 71(6): 15-22.
7. Gargesha, M. and S. Panchanathan, 2002. “A Hybrid Technique for Facial Feature Point Detection”, IEEE
Proceedings of SSIAI, pp: 134-138.
8. HayetBoughrara, Mohamed Chtourou, Chokri Ben Amar and Liming Chen, 2014. “MLP Neural Network
Using Modified Constructive Training Algorithm: Application to Face Recognition”, IEEE
International Image Processing Application and System Conference, pp: 1-6.
9. DewiAgushinta, R., Adang Suhendra, Sarifuddin Madenda and H.S. Suryadi, 2011. “Face Component
Extraction Using Segmentation Method on Face Recognition System”, Journal of Emerging Trends in
Computing and Information Sciences, pp: 67-72.
10. Yi-Qing Wan, 2014. “An Analysis of the Viola-Jones Face Detection Algorithm”, Image Processing On-
Line, pp: 128-148.
11. Zhao, W., R. Chellappa, P.J. Phillips and A. Rosenfeld, 2002. “Face recognition: a literature survey,”
Technical Report CAR-TR-948, Centre for Automation Research, University of Maryland.
12. Ranjana Sikarwar, Arun Agrawal and Shivpratap Singh Kushwah, “A Hybrid Approach to Face Detection
and Feature Extraction”, IOSR Journal of Computer Engineering, 17(2): 78-82.
13. Ijaz Khan, Hadi Abdullah and Mohamed Shamian Bin Zainal, ‘Efficient Eyes and Mouth Detection
Algorithm Using Combination of Viola Jones and Skin Colour Pixel Detection’, International Journal of
Engineering and Applied Sciences, 3(4): 51-60.
14. PriyankaDhoke and M.P. Parsai, “A MATLAB Based Face Recognition Using PCA with Back Propagation
Neural network” International Journal of Innovative Research in computer and Communication
Engineering, 8(2): 89-95.
15. Rafael, C. Gonzalez, Richard E. Woods and Steven L. Eddins, ‘Digital Image Processing Using MATLAB’
by Pearson Education.