See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/371830946

Automated Attendance Monitoring system using discriminative Local Binary

Histograms and PostgreSQL

Article · November 2020

CITATIONS READS

23 15

4 authors, including:

Rupesh Chandrakant Jaiswal

Pune Institute Of Computer Technology
89 PUBLICATIONS 995 CITATIONS

SEE PROFILE

All content following this page was uploaded by Rupesh Chandrakant Jaiswal on 24 June 2023.

The user has requested enhancement of the downloaded file.

© 2020 JETIR November 2020, Volume 7, Issue 11 www.jetir.org (ISSN-2349-5162)

Automated Attendance Monitoring system using

discriminative Local Binary Histograms and
PostgreSQL
1
Akshat Kaushik, 2Yash Gupta, 3Yash Deshpande, 4Dr. R.C. Jaiswal
Department of Electronics and Telecommunications,
Pune Institute of Computer Technology, Pune, India.

Abstract: The aim of this project is to design and implement an attendance monitoring system using Facial Recognition. The
automatic attendance management will replace the manual method, which is time consuming. There are many biometric processes,
in that face recognition is the best method. In this paper, an attendance monitoring system for administrative purposes is created.
In this method the USB camera is fixed at a place and it will capture the image, the face of the person is detected, trained and then
it is recognized with the database and finally the attendance is marked. There are various methods for comparing the faces. OpenCV
(Open Source Computer Vision) is a popular computer vision library which works on the concept of real time image processing and
is supported with various open source implementations of the latest computer vision algorithms. OpenCV 2.7 comes with a pre-
defined programming interface with Python. The Linear Binary Pattern Histograms (LBPH) Algorithm has been used in the
proposed system for facial recognition. The database which records attendance is created using PostgreSQL.

I. INTRODUCTION

Face recognition is generally an uncomplicated task for us humans. According to various experiments young babies of even
a few days old are able to distinguish between faces. Shouldn’t be that complicated of a procedure for a machine then?
Surprisingly, little is known about facial recognition till now. For a successful recognition, are the features within like eyes,
nose etc or are outer features like hairline, head shape more pertinent towards this? The question of how this was analyzed
or how the brain encodes this in general is rather interesting.
There are specialized nerve cells within the brain which correspond to specific localized features of an image, angle and
even movement, as shown by neurophysiologist David Hubel. The visual cortex of the brain should inherently combine the
various sources of information to meaningful patterns, since what is seen is not a scattered reflection.
The main aim of automated facial recognition is to extract the valuable pieces of an image, arrange it into a useful
representation and classify them into distinguishable categories.
The LBPH face recognizer does a similar kind of work. Local Binary Histogram Patterns labels the pixels of an image by
thresholding the surrounding of each pixel and takes into consideration the result as a binary number which is in turn
extremely simple but efficient.

LITERATURE SURVEY

Multiple Face Recognition algorithms are currently existent each of which follow a different approach or methodology and
thereby have different levels of accuracy and different success rates. Role of machine learning algorithms is vital nowadays
for handling the huge data and related accurate and fast classification and recognition [12-17].Although multiple algorithms
might exist to serve this purpose, the algorithm that forms the basis of research and further development in this field is the
Principle Component Analysis (PCA) based on Eigenvalues. This algorithm although being the most inefficient among the
rest, formed the basis of research and led to the development of more efficient and accurate algorithms leading to higher
quality results. But as more and more algorithms were developed that yielded results with higher efficiency PCA was not
used further for the applications of this domain. The Linear Binary Pattern Histograms Algorithm has been used in the
proposed system. The system follows a simple chronological execution of various steps that defines the capability and the
efficiency of the system which includes determining, detecting, capturing of the image in initial stages which is followed
by processing and training the image and finally taking the real-time snapshot and comparing this image with the previously
JETIR2011143 Journal of Emerging Technologies and Innovative Research (JETIR) www.jetir.org 80
© 2020 JETIR November 2020, Volume 7, Issue 11 www.jetir.org (ISSN-2349-5162)
processed and trained set of images. After recognizing the face, the attendance of the particular individual is marked. To
probe into deep, the actual working of the system which focuses mainly on Local Binary Pattern Histograms (LBPH)
Algorithm proves to be a key aspect that forms the basis of the face recognition part of the system[7].

II. PROPOSED WORK

A. Image Acquisition

Scanning a photograph or using an optical camera to capture a live picture of the subject in question are a few
ways an image can be acquired. Another source of face images can also be video which is a multiple images in
quick succession. When the images are particularly of low quality which are in the case of low lighting, various
poses, awkward expressions or illumination, the rate of recognition can be adversely affected in the process.

B. Image Preprocessing

The significant amount of variations in illumination and lighting between gallery and probe pictures is majorly
dealt with various face recognition algorithms. The image pre-processing algorithms that are used to compensate
for such variations in images before the actual recognition is used for the same reason.
The aim of digital image processing is to improve the image data (features) by suppressing unwanted distortions
and/or enhancement of some important image features so that our AI-Computer Vision models can benefit from
this improved data to work on.

Fig. 3.1: The Face recognition overview diagram of our approach.

JETIR2011143 Journal of Emerging Technologies and Innovative Research (JETIR) www.jetir.org 81

© 2020 JETIR November 2020, Volume 7, Issue 11 www.jetir.org (ISSN-2349-5162)

C. Face Detection
Face detection is a technology used in computer vision to understand faces and their location in a 2-Dimension
image. It focuses on the features like eyes, eye-brows, nose, lips and jaw of the face and ignores the rest. Face
detection can be thought of as synonymous to object detection with a particular class in consideration. Object
detection itself is a major task in computer vision which makes face detection a competitive problem to solve.
The coordinates of the faces in the frame are to be extracted from an image with the use of a synthesized,
algorithmic computer program. An overall notion of how a face looks like is used by the program to detect faces
from a given frame.

D. Face Extraction
This module takes care of extracting meaningful information regarding a detected face and devises a feature
vector for the same. Aim of Face extraction is to draw out the relevant information from the provided image.
Holistic feature category and the local features category can be the two classifications in which feature extraction
can be categorized into. The descriptor local binary pattern is used to compare all the pixels including the center
pixel with the neighboring pixels in the kernel to improve the robustness against the illumination variation [6] .
While the holistic feature category comprises of taking the whole face into consideration, Local feature based
techniques try to locate facial features like eyes, nose and mouth which are special to a human face and the known
distances between them are analyzed.

Fig. 3.2: The Face analysis flowchart.

JETIR2011143 Journal of Emerging Technologies and Innovative Research (JETIR) www.jetir.org 82

© 2020 JETIR November 2020, Volume 7, Issue 11 www.jetir.org (ISSN-2349-5162)

E. Training Database and declaring a match

It is required to train the neural network by passing various data sets and creating a large trained database for
high efficiency. After the neural network is trained, it is used on a testing data set to check the accuracy of the
NN. After numerous such iterations, the final model is used in the testing environment of new faces. A database
of known faces is stored and keeps updating regularly. Finally compare the results of the neural network with
those in a database of known faces.
In the attendance monitoring scenario, the detected face is checked and compared against each entry in the
database to find out a match. This will result in the correct face recognition of the students in the classes.

IV. RESULTS AND ANALYSIS

Fig. 4.1: Our System sample screen.

Figure 4.1 shows a sample screen of the proposed attendance monitoring system. The student is automatically
marked absent or present and the status is displayed in real time. Manual marking feature is also provided which
may be used to resolve emergency discrepancies. The marked students are automatically put in the live database
making the system highly automatic. The live class capacity and students who are absent/present for a certain
number of lectures can be viewed by executing simple queries.

The face recognition model correctly detects all the student faces according to the algorithm with an accuracy of
73% achieved on a validation dataset of 800 students from the institution.
The model was also tested on adverse light conditions where ambient light was not sufficient. In such cases the
model performed well with an accuracy of 57%. Figure 4.2 illustrates the obtained results in dim light outside the
classrooms.

The scale of the system was also tested with around 800 students at the same time with 10 simultaneous attendance
systems working concurrently to ingest data of 800 students into the database with an average latency of 50ms.
The live database was tested with complex join queries and the maximum latency observed was 102 ms.

JETIR2011143 Journal of Emerging Technologies and Innovative Research (JETIR) www.jetir.org 83

© 2020 JETIR November 2020, Volume 7, Issue 11 www.jetir.org (ISSN-2349-5162)

Fig. 4.2: Results of our model in dim light.

The attendance system proved to recognize images of indifferent angle and light conditions. The faces which are
not in our training dataset are marked as unknown. The Attendance of recognizing images of students is marked
in real time. And import to excel sheet and save by the system automatically. The accuracy of 73% was achieved
through disruptive local binary histogram patterns and the result has been varying with distance and different
lighting conditions.

Fig. 4.3: Plot of Model Accuracy on Train Datasets.

The results of accuracy comparison for EigenFaces, Fisherfaces and LBPH were important in deciding which
technique to use in the implementation of the attendance management system. Figure 4.3 shows accuracy vs
number of training faces plot results which gave a clear direction and validation for the decision to choose LBPH
for facial analysis. The eigenface algorithm, another method which primarily focuses on linearly projecting the
image space to a low dimensional subspace, which has relatively equal computational requirements. Although,
incisive experimental results show that the other fisherface method has error rates that are significantly lower
than those of the eigenface technique for the comparative assessments done on the Harvard and Yale face

JETIR2011143 Journal of Emerging Technologies and Innovative Research (JETIR) www.jetir.org 84

© 2020 JETIR November 2020, Volume 7, Issue 11 www.jetir.org (ISSN-2349-5162)

databases [4]. For the mentioned reasons LBPH provides a substantial method to use in terms of extraction of
features is concerned.

V. CONCLUSION

The proposed application of face recognition using LBPH proved to be very successful in developing an
attendance management system. Smart attendance management system is designed to solve the issues of existing
manual systems. Face recognition concepts are used to mark the attendance of students and make the system
better. The system performs satisfactorily with an overall accuracy of 73% in different poses and variations. The
database performance was also tested which resulted in good latency and a scalable solution. The system was
tested in a real-life environment in PICT on a total of 800 students. Since the tests are taken in general
uncontrolled environments, the illumination and pose could, to a quite an extent, affect the accuracy rate. The
filtering system introduced reduced these effects. To increase accuracy, a pose tolerant face recognition approach
may also be used [11].

For further improvement in the coming times, the system often faces difficulty in recognizing subjects from a
distance which is compounded with limitations in process power. Better processing provides a more efficient and
effective performance. Although, with the limitations on one side of recognizing a large amount of facial
structures, there is a lot of scope for further improvement. As in this project, a model based approach has been
used, given the opportunity to improve various other models till a required detection and desirable recognition
efficiency is achieved. Enhancement can be in terms of rate of detection, efficiency, precision to more number of
students for an auditorium than a class of 40 students, for which various improvements in algorithms can be
made.

ACKNOWLEDGEMENT

We would express our gratitude towards our mentor, Dr. R. C. Jaiswal for being of great support and guiding me
through the research. He gave this paper the insight and the expertise it needed for making it a presentable one.
His advice, professional acumen, and encouragement proved to be valuable guidance.

REFERENCES

[1] Ahonen, T., Hadid, A. u. Pietikäinen, M.: Face Recognition with Local Binary Patterns. In: European
Conference on Computer Vision 2004, Springer-Verlag Berlin, pp. 469–481
[2] Arubas, E.: Face Detection and Recognition (Theory and Practice), 2013. http://eyalarubas.com/face-
detection-and-recognition.html, accessed on: 16.02.2017
[3] Baggio, D. L.: Mastering OpenCV with practical computer vision projects. Step-by-step tutorials to solve
common real-world computer vision problems for desktop or mobile, from augmented reality and number
plate recognition to face recognition and 3D head tracking. Birmingham: Packt Publ 2012
[4] Belhumeur, P. N., Hespanha, J. P. u. Kriegman, D. J.: Eigenfaces vs. Fisherfaces: Recognition Using
Class Specific Linear Projection. In: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND
MACHINE INTELLIGENCE, VOL. 19, NO. 7, JULY 1997
[5] Bergh, T. F., Hafizovic, I. u. Holm, S.: Multi-speaker voice activity detection using a camera-assisted
microphone array. IWSSIP Bratislava 2016. The 23rd 1
[6] Ashwini, C., et al. "An Efficient Attendance System UsingLocal Binary Pattern and Local
DirectionalPattern." Journal of Network Communications and Emerging Technologies (JNCET) www.
jncet. org 8.4 (2018).
JETIR2011143 Journal of Emerging Technologies and Innovative Research (JETIR) www.jetir.org 85
© 2020 JETIR November 2020, Volume 7, Issue 11 www.jetir.org (ISSN-2349-5162)

[7] Karnalim, Oscar, et al. "Face-face at classroom environment: Dataset and exploration." 2018 Eighth
International Conference on Image Processing Theory, Tools and Applications (IPTA). IEEE, 2018.
[8] Mian, Ajmal. "Realtime face detection and tracking using a single pan, tilt, zoom camera." 2008 23rd
International Conference Image and Vision Computing New Zealand. IEEE, 2008.
[9] Mehta, Preeti, and Pankaj Tomar. "An Efficient Attendance Management System based on Face
Recognition using Matlab and Raspberry Pi 2." International Journal of Engineering Technology Science
and Research IJETSR 3.5 (2016): 71-78.
[10] Bhattacharya, Shubhobrata, et al. "Smart Attendance Monitoring System (SAMS): A Face Recognition
Based Attendance System for Classroom Environment." 2018 IEEE 18th International Conference on
Advanced Learning Technologies (ICALT). IEEE, 2018.
[11] Samet, Refik, and Muhammed Tanriverdi. "Face recognition-based mobile automatic classroom
attendance management system." 2017 International Conference on Cyberworlds (CW). IEEE, 2017.
[12] Jaiswal R.C. and Lokhande S.D, “A Novel Approach for Real Time Internet Traffic Classification”, ICTACT
Journal on Communication Technology, September 2015, volume: 06, issue: 03, pp. 1160-1166.(Print: ISSN: 0976-
0091, Online ISSN:2229-6948 (Impact Factor: 0.789 in 2015).
[13] Jaiswal R.C. and Lokhande S.D “Measurement, Modeling and Analysis of HTTP Web Traffic”, IMCIET-
International Multi Conference on Innovations in Engineering and Technology-ICCC-International Conference on
Communication and Computing -2014, PP-242-258, ISBN:9789351072690, VVIT, Bangalore.
[14] Jaiswal R.C. and Lokhande S.D, “Comparative Analysis using Bagging, LogitBoost and Rotation Forest Machine
Learning Algorithms for Real Time Internet Traffic Classification”, IMCIP-International Multi Conference on
Information Processing –ICDMW- International Conference on Data Mining and Warehousing-2014, PP113-124,
ISBN: 9789351072539, University Visvesvaraya College of Engg. Department of Computer Science and
Engineering Bangalore University, Bangalore.
[15] Jaiswal R.C. and Lokhande S.D, “Statistical Features Processing Based Real Time Internet Traffic Recognition and
Comparative Study of Six Machine Learning Techniques”, IMCIP- International Multi Conference on Information
Processing-(ICCN- International Conference on Communication Networks-2014, PP-120-129, ISBN:
9789351072515, University Visvesvaraya College of Engg. Department of Computer Science and Engineering
Bangalore University, Bangalore.
[16] Jaiswal R.C. and Lokhande S.D, “Analysis of Early Traffic Processing and Comparison of Machine Learning
Algorithms for Real Time Internet Traffic Identification Using Statistical Approach ”, ICACNI-2014-International
Conference on Advanced Computing, Networking, and Informatics),Kolkata, India,DOI: 10.1007/978-3-319-
07350-7_64, Volume 28 of the book series Smart Innovation, Systems and Technologies (SIST),Page:577-587.
[17] Jaiswal R.C. and Lokhande S.D, “Machine Learning Based Internet Traffic Recognition with Statistical
Approach”, INDICON-2013-IIT BOMBAY IEEE CONFERENCE. INSPEC Accession Number: 14062512,
DOI: 10.1109/INDCON.2013.6726074.

JETIR2011143
View publication stats
Journal of Emerging Technologies and Innovative Research (JETIR) www.jetir.org 86