Analou P.
Villanueva
Engr. Alberto Baacia, MEngg
Engr. Rachel M. Chong, PhD
Advisers
Introduction
The Problem
General Objective
Specific Objectives
Significance of the Study
Scope and Limitations
Related Studies
Methodology
Results and Discussion
Recommendation
School Uniform: Benefits
Preventing gang colors
Decreasing of violence and theft
Instilling discipline among students
Reducing the need for the administrators and
faculty to be clothes police
Reducing distractions for students
Instilling a sense of community
Helping schools recognize those who do not
belong on campus
Cebu Institute of Technology University
(CIT-U)
requires the students to wear the prescribed
school uniforms on all schooldays except
Wednesdays and Saturdays
has different sets of uniforms for college
students depending on their course
monitors the students compliance through
manual inspection by the security guards and
faculty members
CIT-U Security Systems
Security guards
RFID readers
Thumb mark readers
Surveillance cameras
Surveillance Cameras
monitor individuals who are coming in and out
the school premises
used only for recording video sequences
can have increased functionality by the
incorporation of image processing algorithms
Image Processing
a technique in which various mathematical
operations are applied to the data (image)
Categories of Techniques:
create an enhanced image that is more useful or
pleasing to a human observer
perform some of the interpretation and
recognition tasks usually performed by humans
CIT-U security systems for the general
population:
Security guards very busy of checking cars,
students, faculty members, staffs and visitors
entering the school premises
Surveillance cameras passive utilization can be
considered as wasteful of resources since the videos
are only stored and analyzed manually
to
add functionality to a surveillance system
by creating an image processing software
that will detect whether the students are
wearing the prescribed school uniform
develop an image processing algorithm that
can detect and evaluate the persons clothes
in video sequences
impart in the developed system the
resulting detection of the persons clothes
evaluate
the accuracy and speed of the
developed system
very useful in schools, universities, private
offices, and even in government facilities
wherein uniforms are required
Surveillance cameras are not only used to
capture video streams but also to detect persons
who are not wearing their uniforms
the resources that are allocated for the
surveillance system will be maximized
can also assist the security guards
the
prescribed school uniforms worn by
male and female students of the College of
Engineering and Architecture (CEA) in
CIT U
for male students
includes light brown
polo, dark brown
slacks, and closed
black shoes
for female
students white
blouse, checkered
slacks and closed
black shoes
Controlled environment
image acquisition
will be done indoor
using a digital
camera
image sequences
will contain one
person only going
towards the digital
camera
Cameras specifications
Resolution 1280 x 720 pixels
Speed of acquisition 30 frames/second
ISO speed Auto ISO
Position - 10 below the imaginary horizontal
line facing the object
Maximum distance from the object 429
Cameras position
429"
Actual physical set-up
38"
30.75"
Actual physical set-up
Camera
Maximum distance:
429
429
Starting point
Distance in 286
Camera
286"
Starting point
Distance in 143
Camera
143"
Starting point
Actual implementation
Software: Scilab and OpenCV
Hardware: Dell Laptop Computer
Specifications:
Memory 1696 MB
Processor i3
Image Classes: 4 classes 1.) male in uniform
2.) male in civilian
3.) female in uniform
4.) female in civilian
Computer Vision: Richard Szeliski
enumerated a wide variety of real-world
applications which include :
retail object recognition for automated
checkout lanes (Figure 1)
Figure 1 http://www.cognitens.com
R. Szeliski, Computer Vision: Algorithms and Applications, September
2010
surveillance monitoring for intruders, analyzing
highway traffic and many other applications
(Figure 2)
Figure 2 http://www.mobileye.com
R. Szeliski, Computer Vision: Algorithms and Applications, September
2010
Image Processing: Debajyoti Karmaker et al
categorized the image processing into compression,
image segmentation, image enhancement, image
filtering, image distortion, image display and coloring
considered pixels as very significant in an image which
are needed in many algorithms in image processing
outlined the three types of processes which the pixels
of an image undergone
D. Karmaker, H. Rahman, M. S. Rahaman, M. S. Rahman, Global Motion
Tracking Six Parameter Model ARPN Journal of Systems and Software, Vol.1
No. 5, Aug 2011.
Images in Green Component: Remi Jean
presented that the digital color cameras generally
use a Bayer mask
presented the filter pattern which is 25% red, 50%
green and 25% blue
based from Bryce Bayers patent used twice as many
green elements to copy the physiology of the human
retina which is most sensitive to green light
Remi Jean, Demosaicing with the Bayer Pattern,
Edge Detection: Indrajeet Kumar et al
presented edge detection as one of the most important
techniques that have been commonly implemented in image
processing
compared the advantages and disadvantages of the various
edge detectors
presented Sobel edge detection exhibits best performance
under noisy conditions as gradient-based algorithms
Indrajeet Kumar, Jyoti Rawat and Dr. H.S. Bhadauria, A Conventional Study
of Edge Detection Technique in Digital Image Processing International Journal
of Computer Science and Mobile Computing , Vol. 3, Issue 4, page 328-334, April
2014
Sobel Edge Detection: Daniel Kim
investigated the strengths and weaknesses of Sobel
Operator and Canny Edge Operator
Sobel Edge Detection: S. Gupta et al
presented the smoothing effect of Sobel to the
random noise of the image
Daniel Kim , Sobel Operator and Canny Edge Detector , 2013
Samta Gupta, and Susmita Ghosh Mazumdar, Sobel Edge Detection
Algorithm, International Journal of Computer Science and Management
Research, Vol. 2, Issue 2, Feb 2013
Background Subtraction: D Stalin Alex et al
presented the background subtraction as widely
used algorithm for detecting the moving objects
from static cameras
presented the main advantage: easy to implement,
fast, effective detection and can provide the
complete feature of the target
D Stalin Alex and Dr. Amitabh Wahi, BSFD: Background Subtraction Frame
Difference Algorithm for Moving Object Detection and Extraction, Journal of
Theoretical and Applied Information Technology, Vol. 60, No. 3, Feb. 28 2014
Edge Thinning: Kamaljeet Kaur et al
considered edge thinning as one of the most
frequently used methods to know the geometrical
feature of the objects
presented the edge thinning applied in the
preprocessing stage to analyze higher level
recognition that preserves the edge information
Kamaljeet Kaur and Mukesh Sharma, A Method in Binary Image
Thinning using Gradient and Watershed Algorithm, International Journal on
Advanced Research in Computer Science and Software Engineering, Vol. 3,
Issue 1, January 2013
Erosion:
https://www.cs.auckland.ac.nz/courses/compsci773s1c/lectures/Im
ageProcessing-html/topic4.htm
considered erosion as one of the algorithms that
removes the imperfections due to noise
introduced the fundamental operations of erosion
https://www.cs.auckland.ac.nz/courses/compsci773s1c/lectures/Ima
geProcessing-html/topic4.htm
Feature Extraction: Dong ping Tian
introduced that color is one of the most important
features of images
proposed a number of important color features like
color histogram, color moments, color coherence
vector and color correlogram
introduced that color moments is one of the simplest
yet very effective features
Dong ping Tian, A Review on Image Feature and Representation
Techniques, International Journal of Multimedia and Ubiquitous
Engineering, Vol. 8, No. 4, July 2013
Support Vector Machine (SVM): Vojtech Franc et al
presented that SVM is one of the most used and best
studied machine learning models
capable of giving high performance classification
accuracy as compared to other classifiers
stated that SVM avoids attributes in greater numeric
ranges and some numerical difficulties during
calculations
Vojtech Franc, Alex Zien and Bernhard Scholkopf, Support Vector
Machine as Probalistic Models, 2011
Object Recognition: Lim Siong Boon et al
proposed method to track human movement using
the color of humans clothing
demonstrated the potential of camera in capturing
important aspects of the objects such as its color,
motion, texture and shape
concluded that the color has more advantages in
tracking application
Lim Sion Boon and Dr Chua Chin Seng, Tracking and Surveillance Urop
Report, Nanyang Technology University, 2005
Object Recognition: M. Marzouk
proposed a modification to the background
subtraction algorithm for motion detection
surveillance system
proposed modification gave a better accurracy by
removing the unwanted objects in the frames in
order to increase the amount of information gained
M. Marzouk Modified Background Subtraction Algorithm for Motion
Detection in Surveillance Systems, Journal of American Arabic Academy for
Sciences and Technology, 1(2): 112-123, 2010
Object Recognition: R. Muralidharan et al
presented SVM based object recognition using the
moment invariant features
proposed method is implemented with testing and
training images
concluded that SVM classifier performs well and
provides high recognition rate compared to other
method
R. Muralidharan and Dr. C. Chandrasekar, Object Recognition Using
Support Vector Machine Augmented by RST Invariants, International Journal
of Computer Science Issues, Vol. 8, No. 5, September 2011
Textile Analysis: Esther Hu et al
matching of the input image to those in the database
is based on four main properties of clothing images:
texture, outline, feature and color
matching is accomplished by a scoring system
images with the highest scores are selected and
presented to the users
E. Hsu, C. Paz, and S. Shen, Clothing Image Retrieval for Smart Shopping,
Technical Report, Stanford University, 2011.
Proposed Method
Image Acquisition
Images in Green Component
Edge Detection
Background Subtraction
Edge Thinning
Masking
Erosion
Image Classification
Outline of the Project Image Processing Algorithms
Images Acquisition
Images in Green Component
Edge Detection
Edge Detection
Background Subtraction
First Edge Thinning
Second Edge Thinning
Vertical Masking
Vertical Masking
Vertical Masking
Vertical Masking
Horizontal Masking
Horizontal Masking
Horizontal Masking
Horizontal Masking
Erosion
Subjects in a Controlled Environment
placing the camera in fixed position and fixed
angle
Execution Time
Region of Interest
Execution Time
Computer Specifications:
Processor: Intel(R) Core(TM) i3-2370M CPU @
2.40GHz
Memory: 4096 MB RAM
Operating System: Windows 7 Professional 64-
bit (6.1, Build 7601)
DirectX Version: DirectX 11
Graphics: 1024 MB DDR3
Total Memory: 1696 MB
Execution Time
Total Elapsed Time 185877125280
Number of frames 2851
processed
Total time (nanoseconds) 1.04483E+11
Average Speed 27.28676049 frames/second
Execution Time at Maximum Distance
Execution Time
Total Elapsed Time 305067302533
Number of frames 4626
processed
Total time (nanoseconds) 1.69277E+11
Average Speed 27.32792314 frames/second
Execution Time at 286" Distance from Starting Point
Gender Attire Detection
Subject
= Female = Male U=Uniform C=Civilian =Correct X=Wrong
RESULTS AND Subject 1
Subject 2
C
DISCUSSION Subject 3 C
Subject 4 C
Feature Subject 5
Subject 6
C
Classification Subject 7
Subject 8
U
C
X
Subject 9 U X
Subject 10 U X
Subject 11 U X
Detection at Subject 12 U
143" Distance
Subject 13 U
Subject 14 C
Subject 15 C
Subject 16 C
Subject 17 C X
Subject 18 C X
Subject 19 C
Subject 20 U
Subject 21 U
Accuracy 15/21 71.43%
Gender Attire Detection
Subject
= Female = Male U=Uniform C=Civilian =Correct X=Wrong
RESULTS AND Subject 1
Subject 2
C
DISCUSSION Subject 3 C
Subject 4 C
Feature Subject 5
Subject 6
C
Classification Subject 7
Subject 8
U
C
X
Subject 9 U X
Subject 10 U
Subject 11 U
Detection at
Subject 12 U
Subject 13 U
286" Distance Subject 14 C
Subject 15 C
Subject 16 C
Subject 17 C
Subject 18 C
Subject 19 C
Subject 20 U
Subject 21 U
Accuracy 19/21 90.48%
More subjects to be included in the training set
Fix installation of the camera
Real-time processing
Open environment
Group of subjects
THANK YOU SO MUCH
