Automatic Attendance System Using Face Recognition

Suman Kumar Pradhan

B. Tech (Pursuing)
Dr.Manish Sharma EEE (7th semester)
Assistant Professor (Electrical Department) Roll No.14233
N.IT, Hamirpur (H.P) e-mail: [email protected]

-
Abstract— consume time and the data is not lost until we delete the data.
The attendance of the students to the classroom is very This method is more efficient in these days.
Important task and if taken manually, much time is lost.
There are several automatic methods available for this System Description
Purpose, i.e. biometric assistance All these methods also The system consists of a camera that captures the photographs
Waste time because students have to do a touch of cola of the room and sends it to the image improvement module.
Your thumb on the scanning device. This work describes the once improvement the image comes within the Face Detection
The efficient algorithm that automatically marks attendance and Recognition modules and so the attendance is marked on
Without human intervention this assistance is recorded by the information server. This is shown within the experimental
In this method the camera is fixed in the classroom and it will setup in Figure2. At the time of enrolment templates of face
capture the image, the faces are detected and then it is pictures of individual students are keep within the Face
recognized with the database and finally the attendance is information. Here all the faces square measure detected from
marked. If the attendance is marked as absent the message about the input image and also the rule compares them one by one
the student's absent is send to their parents. There are various with the face information. If any face is recognized the
methods for comparing the faces. The Eigen face is the one of attending is marked on the server from wherever anyone will
the method. Eigen faces is set of Eigen vectors which are used access and use it for various purposes. this method uses a
in computer vision problem of face recognition. protocol for attending. A time table module is additionally
connected with the system that automatically gets the topic,
Introduction class, date and time Teachers are available in the category and
There are many biometric methods available in which the basic simply press a button to start out the attending method and also
concept is the same. One of them is the identification of the system mechanically gets the attending while not even the
fingerprints. In this method, first, the fingerprints of the intensions of scholars and teacher. during this method lots of
individuals are collected and stored in the database of the your time is saved and this can be highly secure method nobody
fingerprint sensor. For this first we have to collect the will mark the attending of other. attending is maintained on the
fingerprint of each individual. This is done only once or when server therefore anyone can access it for it functions like
a new entry must be added to the database. Then, the administration, parents and students themselves. Camera takes
fingerprints obtained are compared with the images in the the photographs continuously to notice and acknowledge all the
database. if the two fingerprints are equal, the assistance is scholars in the room. so as to avoid the false detection we have
marked as present. But this method has some of the a tendency to square measure using the skin classification
disadvantages. It is by this method that students have to wait in technique [19]. Using this technique enhance the potency and
line, which ultimately consumes a lot of work. If once the finger accuracy of the detection method. during this method 1st the
is not properly maintained or if the fingerprint is not recognized skin is classed and then solely skin pixels remains and every
correctly, attendance will be marked as absent. So this method one different pixels in the image square measure set to black,
is not more efficient. The other biometric method available is this greatly enhance the accuracy of face detection method
the detection of the ball of the eye. In this method, an eyeball .Two databases square measure displayed within the
sensor is used. It detects the blinking speed of the eyeball and experimental setup Figure two. Face Database is that the
also detects the location of the iris. In this method, first the assortment of face pictures and extracted features at the time of
eyeball or iris of each individual is stored in the database. In enrollment and also the second attendance information contains
general, the eyeball is not the same for all people. It has a the knowledge concerning the teachers and students and
difference The image obtained from the eyeball is compared conjointly use to mark attending
with the eyeball in the database. If it is the same, the assistance
is marked. But practically it is not possible. As there is a large
number of students in the class, it is not possible to detect the
eyes of each individual. These disadvantages are overcome with
the help of an automated assistance administration that does not
 before going to the next steps. Histogram normalization is used
for contrast enhancement in the spatial domain. Median filter is
used for removal of noise in the image. There are other
techniques like FFT and low pass filter for noise removal and
smoothing of the images but median filter gives good results.
Image Acquisition
Image is acquire from the camera that is connect above the
board. A camera capture image after every two minutes and
send this image to the computer for processing as shown in fig
2

1 2 3 4
Fig -2 (input image)

Histogram Normalization
Color image is converted to grace scale image for increasing
contrast. As shown in fig 3

Fig-1(Experimental setup)

Technical Requirement Fig -3: Histogram Normalization

 Hardware Requirements
 A standalone computer needs to be installed
in the office room where the system is to be Noise Filtering
deployed. in this system use the median filter for the removal of noise and
 Camera must be positioned in the office room other filter like FFT ,low pass filter this also remove the noise
to obtain the snapshots. Optimum. in the input image
 Resolution: 512 by 512 pixels. Skin Classification
 Secondary memory to store all the images it is use for the increasing the efficiency of the face detection
and database. algorithm its related with binary image use the thresholding of
 Software requirements skin colors
 MATLAB Version 8.5.0(R2015a) or Face Detection
In this stage faces are detected by marking the rectangle on the
higher
faces of the student as shown in fig 4. After the detection of
 Windows XP(Service Pack 2) or higher faces from the next step is cropping of each detected face.
System Algorithm
This section describes the software algorithm for the system. Face detection can be regarded as a specific case of object -class
The algorithm consists of the following steps detection. In object-class detection, the task is to find the
locations and sizes of all objects in an image that belong to a
 Image acquisition
given class. Examples include upper torsos, pedestrians, and
 Histogram normalization
cars.
 Noise removal
 Skin classification Face-detection algorithms focus on the detection of frontal
 Face detection human faces. It is analogous to image detection in which the
 Face recognition image of a person is matched bit by bit. Image matches with the
 Attendance image stores in database. Any facial feature changes in the
In the first step image is captured from the camera. There are database will invalidate the matching process.
illumination effects in the captured image because of different The Viola-Jones algorithm is a widely used mechanism for
lighting conditions and some noise which is to be removed object detection. The main property of this algorithm is that
training is slow, but detection is fast. This algorithm uses Haar
basis feature filters, so it does not use multiplications.
The efficiency of the Viola-Jones algorithm can be significantly
increased by first generating the integral image.

The classifier decision is defined as:

The integral image allows integrals for the Haar extractors to be
calculated by adding only four numbers. For example, the
image integral of area ABCD (Fig.1) is calculated as
II(yA,xA) – II(yB,xB) – II(yC,xC) + II(yD,xD).

fm,i is the weighted sum of the 2-D integrals. is the decision

threshold for the i-th feature extractor. αm,i and βm,i are constant
values associated with the i-th feature extractor. θm is the
Detection happens inside a detection window. A minimum and decision threshold for the m-th classifier.
maximum window size is chosen, and for each size a sliding
step size is chosen. Then the detection window is moved across
the image as follows:
1. Set the minimum window size, and sliding step
corresponding to that size.
2. For the chosen window size, slide the window vertically
and horizontally with the same step. At each step, a set
of N face recognition filters is applied. If one filter gives
a positive answer, the face is detected in the current
widow.
3. If the size of the window is the maximum size stop the
procedure. Otherwise increase the size of the window
and corresponding sliding step to the next chosen size
and go to the step 2.
Each face recognition filter (from the set of N filters) contains
a set of cascade-connected classifiers. Each classifier looks at a
rectangular subset of the detection window and determines if it
looks like a face. If it does, the next classifier is applied. If all
classifiers give a positive answer, the filter gives a positive
answer and the face is recognized. Otherwise the next filter in
the set of N filters is run.
Each classifier is composed of Haar feature extractors (weak
classifiers). Each Haar feature is the weighted sum of 2-D
integrals of small rectangular areas attached to each other. The
weights may take values ±1. Fig.2 shows examples of Haar
features relative to the enclosing detection window. Gary areas
have a positive weight and white areas have a negative weight.
Haar feature extractors are scaled with respect to the detection
window size
 more suitable because of its speed. Hence here we are going to
eigen value method to recognize the faces
After the face detection next step is face recognition this can
be done by cropping the detected face and compare with the
database. In this way face of student. Verified one by one and
attendance is marked on the computer screen as shown in fig 7
Eigen Faces
The acceptable method for face recognition is the Eigen face
approach due to its speed, simplicity and learning capability [5].
The eigen face is one that decomposes the face images into a
small set of characteristic feature, which are the principle
components of the initial training set of face image. Each
discrete face can be represented precisely in terms of a linear
combination of the eigen faces. All the faces can be
approximated by using the best eigen faces.
Calculation of Eigen Weight
Let us consider the image face as I(x, y) which is a two-
dimensional N×N array. The size of the image
is 256×256 which is the vector of dimension 65,536. The faces
are being similar in overall configuration, which are not
randomly distributed in the huge image space [4]. The images
are described by a relatively low dimensional subspace.
Let λ be the eigen value associated with the eigenvector X. It is
a property of a matrix. The vector magnitude is changed not the
Fig -4: Face Detection direction when the matrix act on it.
BX=λX
Initially face detection algorithm tested on variety of images
algorithm was applied to detect face in real time video is shown
in fig 5. To calculate of eigen values and eigen vectors
(B−λI)=0
Where I is the n×n Identity matrix.
The nontrivial solution occurs if and only if

det(B−λI)=0

Where det () denotes determinant.

There are n solutions or n roots of the characteristic polynomial

if B is nxn. Thus B satisfies the equation if there are n eigen
values. Where i=1,2,3….n. Assume that the training sets of
images are Γ1,Γ2,…Γm with each image is I (x, y). Let m be
the number of training images and p be x×y then convert each
and every image into set of vectors and new full-size
matrix (m×p).
To find the mean face by:
Fig -5: Real time face detection
To calculate the mean-subtracted face where i=1,2,….m
The algorithm use the technique of increasing the speed of
B=[Φ1,Φ2,….Φm] is the mean-subtracted matrix vector with
algorithm each crop image is assign to a separate thread for the
its size Amp. The vector matrix is reduced by implementing the
recognition purposes
matrix transformations
Where T is the transpose matrix and C is the covariance matrix.
Face and recognition and attendance By using Jacobi method find the eigenvectors, Vmm and eigen
Face recognition is used to identify the detected faces. There values from C matrix. By comparing other method the accuracy
are many methods available for face detection. But the eigen and reliability of Jacobi's method is high [8]. The linear
value method is the more suitable method [2]. This method is
combination of training set image is determined by these . Advantages
vectors to form Eigen faces, Uk is given by  It saves there time and efforts.
 The software stores the faces that are detected and
Where k=1,2,….m automatically marks attendance
The face vectors of each face based on the Eigen face is:  The system is convenient and secure for the user.
 The software can be used for security purposes in
organization and secured zones.
Where k=1,2,…m
The weight form of the vector is
Disadvantages
 It can only detect face from a limited distance
 The system don’t recognized properly in poor light so may
ΩT=[w1w2……wm] give false results.

Future Scope:
1. Currently, the system has reached the accuracy level up to
80% for partial and dense images. It can further be improved to
obtain higher accuracy levels.

2. Further, 2 or more IP cameras can be employed and each

image can be processed separately. The results of these can be
merged to obtain better results and accuracy in denser
classrooms.

Conclusion
In this system we have implemented an attendance system for
a lecture, section or laboratory by which lecturer or teaching
assistant an record student’s attendance. It saves time and
effort, especially if it is a lecture with huge number of
students. The complete system is implemented in MATLAB.
This attendance system shows the use of facial recognition
techniques for the purpose of student attendance and for the
Fig -7: face recognition and attendance further process this record of student can be used in exam
related issues.
SYSTEM FLOW CHART
REFERENCES
Start http://ieeexplore.ieee.org/document/7916753/author
s
M. Turk and A. Pentland, "Eigenfaces for Recognition",
Journal of Cognitive Neuroscience, vol3, no. 1, pp. 71-86,
1991
Image processing
] R.E. Burkard and E. C¸ ela, “Linear Assignment Problems
and Extensions”, In Handbook of Combinatorial Optimization,
Du
Update the Attendance
Z, Pardalos P (eds). Kluwer Academic Publishers: Dordreck,
1999, pp. 75-149.

Fig -8: System flowchart