You are on page 1of 16

ATTENDOR

An Automated Attendance
Solution

PRESENTED BY- ARJUN PALIWAL


LUV BHATNAGAR
JAINENDER CHAUHAN
YOGESH KUMAR
INTRODUCTION
• Attender is an automated attendence solution.

• This project concerns attendance taking in a


room using image recognition .

• Our objective is simple: determine the


number of people in an image of a classroom.
INTRODUCTION CONT……
• The recognition is based almost entirely upon
sampled inner products between two images.

• Our classroom image was taken using a digital


camera, which yields a 2 dimensional array of
color pixels, each of which has values for the
intensity of red, green, and blue present.
• We take images of empty classroom and filled
classroom.

• This image was then compressed using the JPEG


format and loaded into Mat lab, which converts it a
form it can use.

• We use an indexed color format, which gives each


pixel a value between 0 and 255 that corresponds to
the color of the pixel.
• The resolution of our images were 961 by 1281
pixels.

• The number of people sitting in a class can easily


be found by taking the number of chairs in the
room and subtracting the number of empty chairs.

• A visible chair implies an empty seat and no


student or person in that location.
BLOCK DIAGRAM
IMAGE OF AN EMPTY CLASS
HOW IT WORKS ?
• A chair-sized sub image is clipped from the
image of either the full or empty room, which is
then compared to a image of an empty chair by
taking the inner product.

• Then, another image is clipped from the larger


image of the classroom some offset number of
pixels over from the first image, and is
compared in a similar manner.
CONT….

• When the end of the row is reached, it moves down


by the same offset and begins comparing on the next
row

• If the inner product(always positive, due to the nature


of the values in the image matricies) is higher than a
certain threshold, the program records it as an empty
chair, adding the position where the match occurred
to a list of chair positions
BLURRING PROCESS
• Due to the offset used to save computational time,
we employed a "blurring" subprocess which
reduces the negative effects of not having perfect
alignment when comparing with the inner product.

• The blurring is achieved by simply convolving the


image with a small, normalized, averaging function,
and results in an image which looks almost exactly
like the starting image to the human eye.
DECLUSTERING
• We implemented a declustering algorithm
that removed extra recorded hits in a small
area, leaving only the position with the
highest match, i.e. where the chair and image
were aligned closest.

• This algorithm is used in find chair codes.


CONT….
• Once a list of chair positions is created, with
only one hit per chair, we can then find the
length of that list, i.e. the number of chairs.

• When this is done for both the empty room


picture, and the class picture, then those
numbers can be compared and their difference
is the number of people in the class, attendance
has been effectively taken.
ALGORITHMS
• findchairsempty.m -- Algorithm used on the full room

• findchairsfull.m -- Algorithm used on the empty room

• Clip.m -- Helper function to clip out part of a picture


• edgem.m -- Matrices for the blurring function

• Inner.m -- Finds an inner product

• rmrow.m -- cuts a matrix into a smaller matrix


THANK YOU

You might also like