come from reduction of waste product and its recycling process. To achieve all of
these benefits, number of defected gloves has to be reduced and defect found gloves
has to be recycled. Automated defect identifier and remover has to be applied as an
intelligent inspection system since human workers experience in fatigue by over
working.
Automated defect identifier can be achieved by using concept of control
engineering which studies more about controlling and monitoring systems. It helps
to build a system which can identify the defected product and pinpoints the location
of that defected product. There are two approaches to implement such system. The
first approach is a pattern recognition based on readings from sensor. In this case if
the readings doesnt match with the pattern, the product is to be found as defected.
The second approach works based on threshold values, such as if residual value
comes more than threshold value, the product is found as defected. Image
processing, RFID, Morphological operation and other techniques can be used in
order to accomplish first and second approaches. Advantages and disadvantages of
each techniques will be reviewed and appropriate technique and approach will be
chosen.
Once the defected glove is identified, it has to be removed by using sorting
mechanism. Sorting mechanism has to pick that identified glove and place it in
specified coordination. There are several types of mechanism for pick and place
operation, for instance: intermediate ones are flip flop mechanism, rotating disk
mechanism and complicated one are cartesian coordinate robot, scara robot and
spherical robot and delta robot. Complexity of these robots based on their type and
degree of freedom. Proper selection of sorting mechanism will be explained with
calculations and it will be implemented in the system.
spot, hole, yarn, color bleeding and so on. If these defects will not be detected
properly, these will cause massive affect to the manufactory.
Identification of defected gloves and building a sorting mechanism to pick
and place operation is one the demanding application in industries. This type of
systems will give socially, economically and environmentally benefits and hence it
is complex work to achieve it. According to research done by Karunamoorthy et al.,
(2015) autonomous defect detection system increases such examination efficiency
and propose satisfying recognition precisions for the quality control in industries.
Automated inspection systems have been studied for a long period of time.
Numerous researches have been conducted in this field, and research papers have
been published. Several techniques have been applied and their results have been
learnt, however these systems still have drawbacks in recognizing defects. Tamnun,
Farjana & Ahmed, (2013) also explain the drawback of Artificial Neural Network
(ANN) systems which are designed for finding the defects. They say that artificial
neural network consists of multilayers to identify the defects on the product. ANN is
trained by using back propagation method and the system requires multiple times of
training in order to give better results. Computing time to train the system for
couple of times will require long period of time.
Only several journals have been done on automated glove defect identifier
and remover. Since Asia Pacific University is offering an opportunity to conduct
this research and researcher has chosen this topic based on interest. Aim of this
project is to set to design a system that will identify defects of the gloves and
categorize them into several sorts.
Objectives:
Objectives help to narrow down works to be done in project and these
statements are main objectives of developing project.
1. Design an inspection system to detect defected gloves.
2. Develop a mechanism to categorize gloves into several types.
3. Evaluate the accuracy of detection system.
1.7 Summary
First chapter gave a general introductory to the report. Manual defect
detection technique and its disadvantages were explained. Advantages and
sustainability development factors of automated defect detection system was
explained. Different applications of automatic defect detection for various industries
were given as an example.
It is said that quality control and reduction of wasted materials helps to
improve sustainability. Design of the system is the combination of electrical and
mechanical components and it is mentioned in project background. Research
questions were identified and aim and objectives were set. Justification for this
research were given and essentiality of this application were explained deeply.
Overall, automated glove defect detection system is expected to give
economic and hygienic solution for industries while increasing health and safety
conditions and reducing labor cost. System will able to identify defects on gloves
and sort them based on algorithm. Suitable sorting mechanism will be implemented
and overall system will be tested.
CHAPTER 2
LITERATURE REVIEW
2.1 Introduction
Main aim of doing literature review is to gain related knowledge on
automated glove defect identifier and remover system based on current and past
researches done. By doing literature review, advantages and disadvantages of past
researches will be investigated. Relevant information gained will be used to develop
better automated glove defect detection system.
Nowadays, numerous efforts with diverse procedures and several methods
have been taken into consideration. Since this project is considered as a high
demand project, several researchers have been investigated on similar projects. And
researchers were attained estimated results. But in the other hand, some of the
researchers failed to succeed to build automated defect detection system.
2.2 Literature review
2.2.1 Defect detection techniques using Artificial Neural Network
Researchers (Karunamoorthy et al., 2015) were able to create a system
which can detect defects of the fabric materials automatically by using image
processing in MATLAB. The process is called as Identification of Defective objects
(ID) and it begins by taking an image input to the system, and image will be tested
respected to sample images in database. Image will be enhanced using histogram
function and it will be converted into binary image then binary image goes to
trained neural network classifier. Classifier determines whether fabric is defected or
not and gives a proper command to microcontroller.
Database consists of 256 by 256 defective and defect free fabric images and
this is divided into three groups: dot, star and box patterned fabrics. Results from 20
samples tabulated in table above. ANN providing an astonishing 95% of accuracy
with classifying defected and defect free fabrics while ID method without ANN
defects in fabric material only.
6
ID Method
Without ANN
Percentage
95%
95%
Classification
Not
of Fabric faults
Possible
Accuracy
classification
applicable
Particulars
1
2
3
4
Success rate
Training (%) Testing (%) Overall (%)
100
95
93
100
80
75
100
65
60
100
75
70
Defect free
Hole
Dropped stitch
Scratch
often used in projects and their usage percentage shows less than 12%. ANN was
utilized in their project as a classifier and result achieved is satisfying where
accuracy of the defect detection system reaches to 94%.
2.2.2
database. Image has been converted into gray scale from pixels. Then edge
detection technique has been applied in order to extract the identity of currency.
PCA method is used to decrease the huge dimensionality of the information to
lesser inherent dimensionality of feature scape which is known as independent
variable.
Borkar & Atulkar (2013) conducted their project on developing fuzzy
inference system to detect edges of images using MATLAB software without
defining threshold value. Canny edge detector algorithm has been used in order to
determine the edge of images. Fuzzy logic is a rational classification which is a
postponement of multivalued logic. Fuzzy depended method is a fast and it delivers
reliable correctness and this system is beneficial for image classification. Fuzzy
inference structure is a technique that takes inputs as a vectors and it produces
output based on defined rules. Set rules are built by using toolbox Fuzzy logic in
MATLAB and membership functions are defined.
Researchers (Nagrale & Badge, 2013) worked on project which is for
development of automated inspection system using image processing and their
research differentiates several types of sorting mechanism and explains each
mechanisms advantages and disadvantages. Nagrale & Badge (2013) says that:
machine vision technique has been applied in design automated inspection system
and procedure for this method follows as: thresholds the image where image is
converted into gray scale level and segmentation is done to simplify the
representation of an image for easier examination. Blob technique is then applied in
order to target the object shown in the image. Recognition of pattern has been done
at the end to matching pattern with its template.
2.2.3
pictures, only specific bit is highlighted. Regarding n-bits, plane 0 consists of all
lowest order of bits, and as the number of plane increases, bit number also increases
respectively. Figure 2.2 describes 8 bit image of fabric material. Vital information
about texture of fabric lies on bit plane 0 and bit plane 1. Algorithm is written based
on readings from bit plane 1 and numerous pictures with dissimilar types of defects
were examined. It is found that false alarm rate equal to 2.2% for the examined
samples. Accuracy of finding defects on fabric materials using weighted
morphology reached to 93.2%.
Aziz et al (2013) use morphological process and Discrete Cosine Transform
(DCT) to identify faults on fabric materials. Figure 2.3 above shows the working
principle of inspection system. 1.3MP camera was installed 28cm above the
conveyor belt where the fabric material passes. Image captured from camera was
converted to gray scale and unwanted noises removed by using median filter.
Adjusting the intensity of the image is done by using morphological process. DCT
was used to support morphological process since morphological process doesnt
give acceptable results. By using DCT, high and low frequency components of the
image were found. Components with high frequency of particular image link to
rapidly varying pixel values. In the other hand low frequency components relate to
large measure feature in the picture. In the following step, thresholding process is
11
applied to segment the pixels which are having high value. And defects lies in those
pixels. Several steps done to enhance the quality of the image and defect found
fabric is identified and defect are shown in the monitor itself.
12
and 2 misdetection occurred while testing and their percentages are 2.38% and
4.76% respectively.
Identifying fabric faults using wavelet filter is done by Vaibhav et al., (2013)
and their proposed methodology consist of image acquisition, histogram
equalization, median filter, RGB conversion and wavelet decomposition,
thresholding, morphology filter and finally proposed method can detect the defect.
Image acquired by using CMOS camera and it is changed to gray scale image,
histogram equalizer was utilized to get equally distributed intensity with better
contrast picture. 3x3 Median filter was applied to reduce the Gaussian noise in the
image. Thresholding technique is used to segment the defect. Morphology filter is
used to remove the noise which is created by median filter. Drawback of this
method is wavelet transform fails to identify faults because of color alteration and
smooth edges otherwise it gives promising results.
2.2.4
Sorting mechanisms
Sarkar & Yadaiah (2015) did their project on automation object sorting
which utilizes robotic arm to pick and place operation and states that system
categorizes items based on their color, dimensions, eccentric and sorts using robotic arm
and efficiency reaches 80% using tesseract optical character recognition method. Robotic
arm designed has for small mobile robot and it has 4 axis rotation. These all axis has
been accomplished by using NRS-995 motors and NRS-585 servo motors. When robot
moves up and down, axis 2 and 3 is used to maintain the angle of gripper at constant rate
respect surface. Weight of the load goes up to 250gms and it can grip objects which
is having less than 60mm wide. Arm can reach up to 230mm distance and requires 5
Amp to operate. color objects. Several samples were tested in order to check
accuracy of the system. Accuracy of the system ranges from 60% to 80% based of
the color of the object.
Nagrale & Badge (2013) compares several mechanism for sorting machine
and they are: rotary disk type, pusher type, flip drop. Working principle of rotary
disk type mechanism is that camera fitted on top of conveyor belt. Conveyor belt
runs and transport products and camera captures and sends data to program,
program decides which object to choose and which one to reject and sends
13
14
Year of
publishing name
O
1
2015
2014
Researcher
Methodology used
Outcome
Karunamoort
Identification of defective
95% of
hy et al
accuracy.
Nasira and
samples.
Artificial Neural Network
60% to 75% of
Banumathi
(ANN)
accuracy.
15
Habib et al
Backpropagation is used to
94% of
accuracy.
93.2% of
2012
Priya and
2013
Kumar
Aziz et al
accuracy.
Several defects
are found
Zhang and
(DCT)
Butterworth filter is used and
properly.
92.86% of
Hang
accuracy.
Vaibhav et al
samples.
Median filter and
Promising
results.
8
9
10
2.3
2014
2013
2013
2013
Dhankhar
the defect.
Edge detection technique by
N/A.
2015
and Sahu
Vashishtha
90% of accuary
and Sadim
Principal Component
Borkar and
Analysis.
Fuzzy inference system to
Reliable
Atulkar
correctness.
N/A.
2013
11
2013
Nagrale and
MATLAB
Blob technique is used to
12
2015
Badge
Sarkar and
Yadaiah
accuracy.
13
2013
Nagrale and
mechanism.
Flip flop mechanism is
N/A.
14.
2015
Badge
Phuoc et al
proposed.
3 degree of freedom robot is
Summary
Researches on defect detection method using Artificial Neural Network and
defect detection with other methods have been studied. It is observed that there are
several ways to approach for current development. Image processing using edge
16
detection techniques and sorting mechanisms also have been studied from journals.
Journals gave several dissimilar ideas to solve the problem in designing automated
glove defector and remover. This work does contribute to current industries and
society.
From table 2.3, it can be seen that Artificial Neural Network (ANN) is
mostly used as a classifier for automated inspection systems. Another type of
classifiers used are: Support Vector Machine, Model Based Clustering (MBC) and
Fuzzy Inference System (FIS). ANN, SVM and FIS systems are considered as an
intelligent system where these type of systems can be trained to find defects of
fabric materials. The most effective methods are ANN and it uses tries to reduce the
error by using back-propagation method. However training and finding proper
weights of neurons in ANN using back propagation requires a long period of time
even with high computational power computer. FIS systems does not require a long
period of time for computation and system can work even without training.
Therefore FIS will be chosen and will used as a classifier to the system.
There are also other approaches without using ANN and other intelligent
systems to determine the defects. For example: bit-plane decomposition,
morphological process, Butterworth filter and median filter with thresholding.
Advantages of this systems are that these systems can be created and applied easily.
And they also produce reliable results with more than 85% of accuracy in finding
defects. But these systems are really sensitive to noise where unwanted noise may
lead misleading result. And in addition, any external classifier has to be integrated
in order to determine defected glove. Therefore, these systems will not be utilized in
developing automated glove defect identifier.
Edge detection in image processing will be used for feature extraction. From
the table above, there are several methods to approach edge detection, they are:
edge detection using Canny method, edge detection using Principal Component
Analysis and last two methods are FIS and blob technique to find edges in the
images. Since FIS can detect edges, any other technique to find edge will not be
integrated.
Since title of the project is automated glove defect identifier and remover,
sorting mechanisms are well researched. Sarkar & Yadaiah (2015), build system to
17
sort objects and disadvantage of this proposed system is that accuracy is too low,
where it can sort objects with 60% to 80%. It is aimed to achieve better result. Flip
flop mechanism has been studied and it is known that flip flop system is not
applicable for sorting several types of products. At the end, 3 degree of freedom
robot which is called delta robot has been studied and it is known that it is suitable
for sorting several types of products with fast response. But main disadvantage of
this delta robot is it has a complex structure and it is difficult to troubleshoot when
troubleshooting is required. Concept of sorting mechanisms has been observed and
based on knowledge gain, sorting mechanism will be proposed.
CHAPTER 3
CONCEPT DEVELOPMENT AND EVALUATION
18
3.1
Introduction
This chapter explains about proposed methodology of current developing
prototype based on studies done from previews researchers. Proposed methodology
of this project will be explained with the help of flowchart. Investigation will be
done on choosing proper materials and justification will be given about chosen
component. Design of the concept will be illustrated in 3D and all the necessarily
physics calculations will be done based on fundamental engineering principles.
Aspect of sustainable development of this project in socially, economically and
environmentally manner will be explained. Moreover this chapter of this report will
provide information about ethics and professionalism of the suggested development.
Fabric material glove will be used for this experiment. This type of gloves
are mostly used for safety purposes and also used in cold temperature to maintain
the heat of the hand. It is essential to know the dimension of the fabric type of glove
before determining its defect. Because as mentioned earlier, there are several types
of defects, such as hole, scratch, color bleeding and also wrong sizing.
length
8.3cm
19
23.5cm
Palm width
9.0cm
M
L
21.2cm
22.5cm
8.7cm
9.0cm
24cm
24.5
9.5cm
10cm
Glove with wrong size will be also considered as defected glove and rest
gloves will be sorted based on their proper sizes.
3.2
Proposed Methodology
Since manual inspection is a tedious job for human, an automated glove
defect identifier and remover can be used in manufacturing line to increase the
quality and sustainability of the product. Proposed method has to able to identify
defect of the glove and remove the defect found gloves. Methodology will be
explained with a help of block diagram and flowchart.
This project is going to be implemented based on several engineering
subjects, such as mathematics, dynamics, strength of material, control engineering,
machine design, computer aided programs and mechatronic designs and many
more. By using fundamentals of engineering concepts, automated glove defect and
remover will be completed.
Several journals have been studied in literature review in order to gain
knowledge about automated defect detection systems. Several methods of finding
defects and sorting mechanisms have been analyzed through the journals. Results
show that researchers used novel techniques and methods and they have achieved
high rate of success. Based on their designs and methodologies, current developing
project will consist of main three parts as shown in Figure 3.2 below:
1.
2.
3.
20
21
Gravity pulls the gloves downwards and gloves will go through the glass. Images
will be taken by using two cameras which are located perpendicular to glass and
they will be aligned at the same line. Continuous images will be sent to computer
and analysis on defect detection of particular glove will be carried on. Classifier
will categorize the defected and defect free products and it sends proper command
to microcontroller. Microcontroller controls the movements of motor and motor
moves the sorting mechanism and gloves will be sorting properly.
22
researchers. The reason why current developing sorting is twice smaller is because
of limited space and limited budget.
Flowchart given in Figure 3.5 below explains how the system works in
detail. Program starts and runs the whole system, and input images to the system
will be taken by using two cameras. Program filters the image taken and extracts the
feature needed from the image. Then program checks the glove whether it has a
defect or not. If the defect is found, program sends command to sorting mechanism
to take a glove from particular location and pick it and place it in recycle bin. If the
glove is found as defect free, program checks its type, once the type of the glove is
identified, program classifies and sends command to sort gloves to sorting
mechanism.
23
Figure 3.5: Flowchart of the automatic glove defect detector and remover
24
microcontrollers
are
available
in
the
market
nowadays.
low cost, small and programmable chip which can be utilized in embedded systems.
Microcontroller can be categorized based on their memory structure, instruction set
and width of their bus. Table 3.1 below, compares the specifications of three type of
microcontroller.
Table 3.2: Microcontroller Comparison
Microcontroller types
Arduino Uno
PIC 16F877A
8051 Microcontroller
Pins
20 pins
40 pins
32 pins
Flash Memory
32KB
8KB
32KB
Clock Speed
16MHz
20MHz
24MHz
Output Current
50mA
25mA
30mA
Price
RM 50.00
RM 17.00
RM 20.00
Arduino has 20pins and 14 of them can be used for digital I/O purpose.
However PIC and 8051 microcontrollers have 40pins and 32pins respectively. Flash
memory of the microcontroller is 32KB which is equal to flash memory of 8051
Microcontroller, and 4 times bigger than PIC16F877A microcontroller. Arduino
provides less clock speed but it gives higher output current compares to other
microcontrollers.
Arduino Uno is one the most used microcontroller in embedded systems,
because this type of microcontroller is ready to use. Arduino has built in 5V
regulator, an oscillator, serial communication interface, a micro controller and its
software to compile the programs. And also Arduino consists of big library to
control and monitor external hardware circuitry. PIC microcontrollers also knows as
reliable and percentage of malfunctioning is found very less. Since RISC
architecture is used in PIC, it provides high speed with less power consumption for
controlling and monitoring systems. Disadvantages of PIC family is that program to
written to control the embedded system will be lengthy since it is using 35
instructions. Lastly, 8051 microcontroller is faster than others, while it works
24MHz with 32KB flash memory but its architecture is complicated and which
26
Feature
Raspberry PI
CM1080P
Fixed focus
Webcam
EW-101SR
Crisp VGA resolution
Frame rate
exposure
30fps
Temperature
-30 C to 70 C
Resolution
5M pixel
-40 C to 85 C N/A
1.3M pixel
5M pixel
Price
RM 40.00
RM 105.00
30fps
15fps
RM 18.00
DC gear motor
Lifting
Servo motor
Controlling
Stepper motor
Controlling the
mechanisms
the positions
Position
rotation
Strong torque
control
360
3 wires
RM 24.90
360
6 wires
RM 82.40
Feature
PWM
Degree of rotation
Wire connection
Price
360
2 wires
RM 45.00
DC gear motor is mostly used for lifting mechanism, while stepper motor is
used for application which requires high torque values. Servo motors is more
applicable for controlling the positions in certain application. Rotation of all these
three motors is 360. Pulse width modulation is generated for DC gear motors by
changing the pulse of the signal in microcontroller. Servo motors comes with built
in encoder to control the position precisely. In this application, DC gear motor and
servo motor will be needed and they will be used for conveyor belt and sorting
mechanism respectively.
Conveyor belt
Conveyor belt is widely used in manufacturing line and it is used to
transport products on it. It consists of frame, pulleys, bearings, motors and belts.
28
One side of the conveyor belt is powered by motor and this makes belt rotates.
Table 3.4 below gives comparison between several types of pulleys.
Table 3.5: Comparison of belt drives
Pulley types
Belt type
Material
Groove number
Price
V belt
Bearing Pulley
V-belt
Plastic
1
RM 25.00
V-belt
Round
ABS
N/A
RM 10.00
1
RM 32.00
Bearing pulley will be chosen since it is a low cost pulley with durable
quality.
Bearing is a mechanical component that is utilized to reduce the friction between
relative moving parts. There are some designs based on application, for instance, a
bearing that gives free linear movement to the moving parts or free rotation around
the constrained axis. Main aim of using bearing is to achieve to required output by
minimizing the friction. Number 210 bearing has been chosen in design, because it
6
is found that lifetime of the bearing is 6.710 cycles .
Software
There are mainly two software used by engineers to do their simulation,
design, modelling and prototype testing. These software are known as MATLAB
and LABVIEW. Both software have advantages and disadvantages. MATLAB
provides bigger library data while LABVIEW provides easier access for data
acquisition. With huge library data, it is much easier to do programming with
MATLAB. Therefore MATLAB will be used as a main program and corresponding
GUI will be created. LABVIEW will be interfaced if it is required. Arduino will be
interfaced to control the sorting mechanism.
3.4 Concept design derived from fundamental engineering principles
This chapter explains more about calculations on physical laws which are
going to be applied while building this project. Fundamentals of several engineering
concepts will be utilized in order to achieve the objective.
29
A = width (250mm)
B = thickness (3mm)
Substituting A and B values into equation 3.1:
m=
1.7 kg
x 0.25 m x 0.003 m=0.001275 kg /m .........................................(3.2)
3
m
Result (3.1) shows that mass of the rubber kilogram mass per meter. In following,
required length of the rubber belt for conveyor will be found.
1
2
Lo= ( d l + d s ) +2C + ( d ld s ) ................................................................(3.3)
2
4
Lo = the total length of the belt
d1 = Diameter of large pulley (6cm)
d2 = Diameter of small pulley (6m)
C = Center distance between the two pulleys (60cm)
1
Lo= ( 6 cm+6 cm ) +260 cm+ ( 6 cm6 cm )2=138.85 cm ....................(3.4)
2
4
Total mass of the belt can be found by multiplying the required length of the belt by
mass of the rubber kilogram mass per meter.
m=0.001275
kg
1.3885 m=0.0018 kg ....................................................(3.5)
m
30
d 1d 2
1 6 cm6 cm
= sin
=0 rad
C
60 cm
.....................................................(3.7)
1
=sin
2 r
x n ................................................................................................(3.9)
60
r = radius of the pulley which is 0.6cm
n = speed of the motor and it is assumed to be 20rpm
v=
2 (0.06 m)
x 20=15 x 105 m/ s ..........................................................(3.10)
60 x 1000
T =F x d ....................................................................................................(3.12)
F = force found and it is equal to 2.5N
d = distance known from the axis of revolution and it is assumed to be 50cm
T =F x d=2.5 N0.5 m=1.25 Nm ............................................................(3.14)
Calculation for require torque has been done and it is observed that torque is
1.25Nm including loads. Based on this calculation, DC motor SZGH - 06040DC
will be chosen since SZGH - 06040DC motor can provide 1.5Nm torque.
Calculation for bearing
Number 210 bearing has been chosen in designing of conveyor belt, because
its having 30mm inner radius and 60mm outer radius.
Formula given below is used to find the N:
N=
106 C3
.................................................................................................(3.15)
F3
F = Force is equal to 2.5N
C = 99000 for steel material
N=
L=
3
3
106 C3 10 ( 99000 )
=
=6.3310 9 .......................................................(3.16)
3
3
F
( 2.5 )
N
60 n .....................................................................................................(3.17)
n = speed in rpm, and it is assumed to be 20rpm
1.2210 10
1
N
60
L=
=
=5280000 cycles=5.28106 cycles ...................(3.18)
60 n
20
6
32
33
34
be analyzed for 10 days and half month will be used for improvement of the total
system.
Components cost estimation
Table 3.6: Estimated cost of components
Component
No. of piece
Arduino Uno
45
45
Servo motor
80
160
DC gear motor
52
52
Camera
Material of conveyor belt
Pulley
2
1
2
45
55
10
90
55
Rollers
35
70
Pulley wheels
20
80
Bearings
4
1
1
15
100
90
20
60
100
90
Total: 822
3.8 Summary
Methodology, concept design of the developing prototype has been done in
this chapter. Methodology of the project explains how to build automated glove
defect identifier and remover with help of block diagram and flowchart. Block
diagram is used to give visual overview and it is also used to give proper
explanation on working principle of developing prototype. Investigation on
materials and components have been done to choose suitable component to the
project. By doing investigation, suitable microcontroller, camera, motor, conveyor
belt and software has been chosen. Calculations have been on conveyor belt, since
this project does not have much moving parts. From the calculation, motor with
required torque has been found and it has been selected. Calculation for bearing has
been done to show the total life cycle of the bearing. Sustainable development
report gave an overview about sustainability of the developing project based on
economically, socially and environmentally aspect. Project management has been
done in order to manage the time and money required. Moral professionalism shows
that project is following all the regulations set by Malaysian government. In the
other hand, ethical consideration shows that only certified or trial version of
software have been used while developing the project.
37
REFERENCES
Aziz, A., Haggag,, A. and Sayed, M. (2016). Fabric Defect Detection Algorithm
Using Morphological Processing and DCT. International Journal of Engineering
and Innovative Technology (IJEIT), 4(11), pp.1-5.
38
Borkar, A. and Atulkar, M. (2013). Fuzzy Inference System for Image Processing.
International Journal of Advanced Research in Computer Engineering &
Technology (IJARCET), 2(3), pp.1007-1010.
Dhankhar, P. and Sahu, N. (2013). A Review and Research of Edge Detection
Techniques for Image Segmentation. International Journal of Computer Science
and Mobile Computing, 2(7), pp.86-92.
Habib, T., Hossain, R., M, R. and Ahmed, F. (2014). AUTOMATED FABRIC
DEFECT INSPECTION: A SURVEY OF CLASSIFIERS. International Journal in
Foundations of Computer Science & Technology (IJFCST), 4(1), pp.17-26.
Karunamoorthy, B., Somasundareswari, D. and Sethu, P. (2015). AUTOMATED
PATTERNED FABRIC FAULT DETECTION USING IMAGE PROCESSING
TECHNIQUE IN MATLAB. International Journal of Advanced Research in
Computer Engineering & Technology (IJARCET), 4(1), pp.63-69.
Nagrale, S. and Bagde, S. (2013). Application of Image Processing For
Development of Automated Inspection System. International Journal Of
Computational Engineering Research, 3(3), pp.103-107.
Nasira, M. and Banumathi, P. (2014). Automatic Defect Detection Algorithm for
Woven Fabric using Artificial Neural Network Techniques. International Journal
of Innovative Research in Computer and Communication Engineering, 2(1),
pp.2620-2625.
Phou, T. and Truong, N. (2015). Design and development of the sorting system
based on robot. In: 15th International Conference on Control, Automation and
Systems. Busan: ICROS, pp.1639-1644.
Priya, S., Ashok, T. and Paul, V. (2012). Computer vision. In: Defect Detection In
Wooven Fabric Using Weighted Morphology. Coimbatore: IEEE, pp.1-6.
Sarkar, S. and Yadaiah, K. (2015). Automation of Object Sorting System Using
Pick & Place Robotic Arm & Image Processing. International Journal &
Magazine of Engineering, Technology, Management and Research, 2(8), pp.54-63.
Vaibhav, V. and Bhangale, B. (2015). Fabric Defect Detection using Wavelet Filter.
In: International Conference on Computing Communication Control and
Automation. Chakan Road, Induri. Pune: IEEE, pp.712-716.
39
40
APPENDIX A
LOG SHEET
41
APPENDIX B
DRAFT PROPOSAL
Dear KHUSANKUL(TP029644-UC4F1603),
Thank you for selecting this project proposal from our
FYPBaNKhttp://fypbank.apiit.edu.my/
42
Your supervisor has approved your request to continue working on this project
proposal below.
You may continue working on this project proposal.
Project proposal id
: 1536
Project Title
and remover
Name of Supervisor
: siong.chung@apu.edu.my
Regards,
Customer Support Department, FYPBaNK, APU
43
APPENDIX C
Project Specification form
The following should be highlighted in the project specification /brief: (Form attached
below)
44
A. Project Title.
B. Brief description on project background. (.i.e. problem context, rationale,
description of problem area, nature of challenge)
C. Brief description of project objectives. (i.e. scope of proposal, constructs used,
limitations and significance)
D. Brief description of the system/model/design that will be used in this proposal.
E. Academic research being carried out and other information, techniques being learnt.
(i.e. literature - what are the names of books you are going to read / data sets you
are going to use)
G. Brief description of the evaluation and analysis proposed for this project.
(i.e. project deliverables and hypothesis, correlation test etc)
Project Title.
Automated glove defect identifier and remover
45
B.
This research studies on developing automated glove identifier and remover. By using
automated inspection system, gloves with small holes and cuts, and different thickness
and different width of glove compared to expected results could be taken away by
sorting mechanism. And this results in reduction of defected gloves and it will be
beneficial socially, economically and environmentally. From the social aspect, it is
beneficial for customers and clients. Because they will receive high quality and nondefected gloves. Economically and environmentally benefits come from reduction of
waste product and its recycling process. To achieve all of these benefits, number of
defected gloves has to be reduced and defect found gloves has to be recycled.
46
Automated defect identifier and remover has to be applied as an intelligent inspection
system since human workers experience in fatigue by over working.
Several journals have been studied in literature review in order to gain knowledge about
automated defect detection systems. Several methods of finding defects and sorting
D.
mechanisms
have been of
analyzed
through the journals.
Brief description
the system/model/design
thatResults
will be show
used inthat
thisresearchers
proposal.
used novel techniques and methods and they have achieved high rate of success. Based
on their designs and methodologies, current developing project will consist of main
three parts:
Mechanical design: it will consist of conveyor belt and sorting mechanism.
Sensor to get input: input will be taken by camera and images will be sent to program
for analyses purpose.
Program: controlling process will work based
on algorithm written and location of
47
defected glove will be identified on conveyor belt and corresponding command will be
send to microcontroller which controls the sorting mechanism. And microcontroller will
Nasira & Banumathi (2014) proposed an algorithm to identify and classify defects of
woven fabric materials using ANN. Initially image is acquired and pre-processed and
then its histogram is normalized. These steps were used to extract the feature in the
image. Minimum resolution used was 300dpi, in contrast maximum human vision is
E.
300dpi. Image is stored in the memory and it converted into gray scale level.
Academic research being carried out and other information, techniques being
Resolution
is increased
by 100dpi
by of
step
and you
till it
to read
1200dpi
as sets
a
learnt. (i.e. literature
- what
are the step
names
books
arereaches
going to
/ data
you
are going
to use) By using digital filters, noise in the image is reduced. Filtered
maximum
resolution.
image is then changed into binary scale. By using imhist function in MATLAB,
et al., 2015)
were
able toused
create
a system
histogram ofResearchers
the image is(Karunamoorthy
taken. Back-propagation
neural
network
to train
the which can
detect defects of the fabric materials automatically by using image processing in
ANN classifier.
The process
is called
Identification
Defective
Research on MATLAB.
automatic fabric
fault detection
hasasbeen
proposed byofHabib
et al., objects
(2014) (ID) and it
begins
imageinspection
input to the
system,
and the
image
willand
be selects
tested respected to
and researchers
sayby
thattaking
fabricandefect
firstly
analysis
scene
in database. Image
be enhanced
using
histogram
feature in thesample
scene.images
Then information
goes towill
image
acquisition,
followed
by function
image and it will
converted
binary image
then binary
goes toinspection
trained neural
network
preprocessingbe and
featureinto
extraction.
The following
stepimage
of defect
is
whether
is defected
not and gives a proper
classification,classifier.
and thisClassifier
research determines
mostly focuses
on fabric
different
types of orclassification
techniques. command to microcontroller.
Database
256 by 256
defectiveare:
andArtificial
defect free
fabric
images and this is
Paper compares
four consists
types of ofclassifiers,
classifiers
Neural
Network
divided
into three
groups:
dot, star
and box patterned
Results from 20 samples
(ANN), Support
Vector
Machine
(SVM),
Statistical
Test (ST)fabrics.
and Model-Based
tabulated
in 2.1
tableabove
above.
ANNwhich
providing
an is
astonishing
95%
accuracy with
Clustering (MBC).
Figure
indicates
classifier
often chosen
andof
often
classifying
defected
and
fabrics method
while IDand
method
without ANN
utilized. Classifier
Neural
Network
is defect
one of free
the popular
it is considered
as defects in
fabric material
only. Other three classifier not often used in projects and
most used classifier
with 76.47%.
their usage percentage shows less than 12%. ANN was utilized in their project as a
classifier and result achieved is satisfying where accuracy of the defect detection
system reaches to 94%.
A research on edge detection technique for image segmentation proposed by Dhankhar
& Sahu (2013) states that objective of this study is to deliver a review of several
methods for image processing based on edge detection technique and examining their
performance based on problems such as false edge detection, noisy pictures or lost
edges. Steps for edge detection were given
48 and they are filtration, enhancement and
detection. In the first step filtration, some of the random noises were filtered and image
smoothened, then in following step quality of an image is improved and lastly image is
There are also other approaches without using ANN and other intelligent systems to
determine the defects. For example: bit-plane decomposition, morphological process,
Butterworth filter and median filter with thresholding. Advantages of this systems are
that these systems can be created and applied easily. And they also produce reliable
results with more than 85% of accuracy in finding defects. But these systems are really
sensitive to noise where unwanted noise may lead misleading result. And in addition,
any external classifier has to be integrated in order to determine defected glove.
Therefore, these systems will not be utilized in developing automated glove defect
identifier.
Edge detection in image processing will be used for feature extraction. From the table
above, there are several methods to approach edge detection, they are: edge detection
using Canny method, edge detection using Principal Component Analysis and last two
methods are FIS and blob technique to find edges in the images. Since FIS can detect
edges, any other technique to find edge will not be integrated.
Since title of the project is automated glove defect identifier and remover, sorting
mechanisms are well researched. Sarkar & Yadaiah (2015), build system to sort objects
and disadvantage of this proposed system is that accuracy is too low, where it can sort
F.
objectsBrief
with description
60% to 80%.of
It isthe
aimed
to achieve better result.needed
Flip flopbymechanism
has
materials/methodologies
the proposal.
(i.e.
data
collection
testing
and
been
studied
and itmethods,
is knownequipment,
that flip flop
system
is etc.)
not applicable for sorting several
types of products. At the end, 3 degree of freedom robot which is called delta robot has
been studied and it is known that it is suitable for sorting several types of products with
fast response. But main disadvantage of this delta robot is it has a complex structure and
it is difficult to troubleshoot when troubleshooting is required. Concept of sorting
mechanisms has been observed and based on knowledge gain, sorting mechanism will
be proposed.
49
Structure of automated inspection system will be made by aluminum material which can
be renewed and reused. And also, 3D design is drawn by using help of CAD software
and drawing will be used to give better visualization while building the prototype.
Circuit scheme of the system is simulated using Proteus program before implementation
and it helps to reduce the purchasing unwanted and extra electronic components. System
sorts defected gloves and defect found gloves goes to recycling process and this helps to
improve wastage of fabric materials. Whole system will be powered up by using 12V
rechargeable battery, and it can work for a long period of time without having any
fatigue issues. This makes system more environmental friendly.
As it is known that manufacturing line is relatively hot and humid compared to
temperature outside. Automated glove defect detection system and remover will be used
in that situation and it prevents humans health by taking their tedious daily job. And
G.
description
of the evaluation
analysis
proposed
for this
also itBrief
prevents
human inspectors
eye powerand
since
checking
small defects
on project.
fabric
(i.e. project deliverables and hypothesis, correlation test etc)
reduces power of the eyes. The life span of the automated defect detection system for
glove will be enhanced to reach maximum life span and prototype can be used for a long
period of time.
50
51
H.
environmentally
and
socially
benefits
for
current
generation.
52
APPENDIX D
ETHICS FORM
53
APPENDIX E
COMPONENTS DATA SHEET
54
55
APPENDIX F
TURNITIN REPORT
56