Name :KHAIRUL NAIM BIN ARIFFIN

Matric No. :211061220

student number
(refer attendance
list)
Code/Course : EMJ36203 / Machine Vision
Item : Assignment
Semester/session : 1/2022-2023
Due date for 18 / 01 / 2023
submission

Question1

An industrial automated machine vision inspection system shown in Figure 1 is proposed for Spark
Industries to improve their production rate. The layout of the object of interest is depicted in Figure
2(a), showing the object’s radii respectively. Based on your knowledge of the course, propose a
machine vision system to separate the objects of interest according to their respective classes. Use the
appropriate diagram(s) to assist your answer. Finally, decide to which group the new object given in
Figure 2(b) belongs.

Use the last digit of your matric number as variable M to add to the radius, R of the new object.

For example, if your matric number is 191061235, then R is equal to 5+1 = 6.

If your matric number is 191061230, then R is equal to 0+1 = 1.

Please refer rubric for details.

(20 Marks)

Figure 1

1
 Figure 2

Rubric for Q1 (20/20)

No Criteria Rubric

Excellent Good Fair Unsatisfactory Not available

4 3 2 1 0

1 Idea/concept Excellent idea/concept is Good idea/concept is Fair idea/concept is Unsatisfactory Not present.
[4 Marks] proposed. proposed. proposed. idea/concept is Plagiarism is
The chart(s) or figure(s) proposed. detected.
is/are excellent, and the The chart(s) or The chart(s) or figure(s)
calculations are correct. figure(s) is/are good, is/are acceptable, and the
and the calculations calculations are correct.
are correct.

2 Flow Chart Very clear and accurate Good flow chart is Fair flow chart is Poor flow chart is Not present.
[4 Marks] flow chart is presented presented presented presented. Plagiarism is
detected.

3 Image Good image processing Good image Fair image processing The proposed image Not present.
processing steps are proposed. The processing steps are steps are proposed. The processing steps are Plagiarism is
[4 Marks] explanation is very clear proposed. The explanation is moderate. not relevant. detected.
and accurate. explanation is clear.

4 Feature Good Feature Extraction Good Feature Fair Feature Extraction Feature Extraction Not present.
Extraction method(s) is/are proposed. Extraction method(s) method(s) is/are proposed. method(s) is/are Plagiarism is
[4 Marks] The explanation is very is/are proposed. The The explanation is proposed. The detected.
clear and accurate. explanation is clear. moderate. explanation is weak.

5 Classification Good Classification Good Classification Fair Classification Classification Not present.
[4 Marks] method(s) is/are proposed. method(s) is/are method(s) is/are proposed. method(s) is/are Plagiarism is
The explanation is very proposed. The The explanation is proposed. The detected.
clear and accurate. explanation is clear. moderate. explanation is weak.

2
Question 2
The Health Center of Universiti Malaysia Perlis (UniMAP) is in the process of automating body mass
index (BMI) calculation based on the capturing process of a patient's body. Design a machine vision
system so that the patient's BMI can be easily recognized.
(12 Marks)

IMPORTANT NOTE

• THIS IS NOT AN ESSAY QUESTION. Instead of writing long paragraph, be creative in your presentation. Remember, your
presentation and given details must be easy to understand.

a. VERY IMPORTANT! Marks may be deducted if students write long paragraphs or an essay. Your answer for this
assignment should not be more than 4 pages long.

3
Question 1
Main idea
Industrial automated machine vision inspection system is becoming increasingly popular in the
manufacturing industry. These system uses camera and image processing software to inspects products
for defects and other qualities. They can be used to inspect a wide range of products from food to
electronics and beverage products.

Since the Spark Industries are planned to proposed a machine vision system to separate the objects of
interest according to their respective classes, there are 8 things that should be embedded in the proposed
of machine vision system in order to improve the production rate

System
would need
to be able to
handle a
System would variety of Combination of
be able to different sensors, such as
handle objects materials cameras, lasers,
that are and infrared
moving or sensors, to detect
changing and classify
shape objects

System able to machine vision Sensors would be

handle a variety system to separate able to the detect
of different the shape, size and
lighting
the objects of color of the object
conditions, as well interest according to as well any other
as different levels their respective distinguishing
of noise classes objects

System would
then use a
System need to
classification
be able to
algorithm to
distinguish
assign each
between System would also object to its
similar objects need to be able to respective
handle a variety of classes
different objects,
from small items to
large

1.0 System Algorithm

Based on the research that has been made, the system algorithm that we chose for the vision sensors
to identify and separate object according to their respective classes are the edge detection and pattern
recognition

4
 edge canny edge Pattern
detection detector Recognition

• locate the boundaries • uses a combination of • once the edge is detected

between 2 objects gradient magnitude and it identify the objects
• can be done by using direction based on shape and size
canny edge detector of the image

Figure 3: The edge detection algorithm process

Edge detection algorithms

• Detect edges in various ways such as by looking for changes in brightness, color and texture.
Also detecting the edges in variety of shapes such as circles, squares and triangle

• uses high resolution cameras that capable of capturing images with high degree of accuracy
typically connected to the computer system that is capable of analyzing the images and
detecting any defects

• used in type of inspection systems such as optical character recognition (OCR) systems and
automated optical inspection (AOI) systems

• OCR systems are used to detect text in images while AOI used detect defects in objects.

Pattern recognition algorithms

• ability of the sensors to detect and recognize the pattern by analyzing the visual information
that is captured by the sensor.

• The sensor then will use the information collected to identify objects and classify them
according to their classes

• There are 3 type of pattern recognition in vision sensors which is convolution neural network,
support vector machine, decision trees and k nearest neighbor (KNN’s). Each of this algorithm
has their own strengths and weakness, and can be used to identify different type of objects

2.0 flowchart of edge detection algorithm and pattern recognition algorithm

For a clear explanation of vision sensors that have been proposed we have visualized a flowchart of the
system algorithm edge detection and pattern recognition functions

5
 Gradient Edge Post
start Pre processing Output End
Calculation Detection Processing

Figure 4: flowchart of edge detection algorithm

pattern
collection analyzed pattern inform
start end
of data the data evaluation decisions
or actions

Figure 5: flowchart of pattern recognition algorithm

3.0 Image processing

Edge Detection image processing

Pre • step involves

smoothing the image
to reduce noise and
Processing enhance the edges
• steps involves
calculating the
gradient of the
images intensity
• Canny edge
detector algorithm
Gradient has been proposed
for this system as
Calculation uses two 1-D
convolution kernels
to calculate the
gradient of the
image intensity

• Steps involves
Edge applying a threshold
to the gradient
Detection image to detect the
edges

• steps involves
Post morphological
operations to the
Processing detected edges to
refine them

Output • steps involves in

displaying the edges

6
Pattern Recognition image processing

• remove any noise or

Pre Processing artifacts that may be
present

Feature • identify and extract

extraction feature from the
images that can be
techniques used to identify
patterns or objects
applied

Pattern • identify pattern or

recognition objects within the
image
evaluation

• the results of the

classification can be
Pattern used to make
recognition has decisions or take
actions based on the
been classified identified pattern or
objects

4.0 Feature extraction

Feature extraction in edge detection is a process of extracting meaningful information from an image. It
is used to identify and extract features such as lines, curves, corners and other shape from an image.
From the research, we have chosen Canny’s method as the proposed of feature extraction. This is
because this method is a popular technique used in computer vision and image processing. Canny’s edge
detection algorithm composed of 5 steps:

I. Noise reduction

II. Gradient calculation

III. Non maximum suppression

IV. Double thresholding

V. Edge tracking

7
 non maximum
noise reduction suppression
gradient -used to thin the edge tracking
-reduce the edges in the images double thresholding
amount of noise calculation -connect the
-comparing the -identify strong and
in the image by -applying sobel gradient magnitude strong edges in
weak edges in the
applying operator the of each pixel to its images the images
gaussian filter to image neighbor and
suppress any pixel
the image that are not local
maximum

5.0 image classification

image classification can be divided in two groups which is:

• Supervised learning

• Unsupervised learning
Based on figure 2 the image classification image that been proposed are the supervised learning
classifier which is k- nearest neighbor classifier

K-nearest neighbor classifier method

K= 2
Matric number is 211061220. The last digit of my matric number is 0
R= 0 +1 = 1

Radius R<5 or R>5 RGB color table Color Category

/Classes
2.0 2.0<5 255,255,0 Yellow
2.5 2.5<5 255,255,0 Yellow
3.0 3.0<5 255,255,0 Yellow
4.0 4.0<5 255,255,0 Yellow
9.0 9.0>5 255,0,0 Red
8
 9.5 9.5>5 255,0,0 Red
10.0 10.0>5 255,0,0 Red
10.0 10.0>5 255,0,0 Red
10.5 10.5>5 255,0,0 Red

Since the radius of new object M is below than the radius of 5, then the new object will be classified as
Yellow classes

9
Question 2

determine the specific Improve accuracy and

requirements for the system effieciency of the system
regular system updates
-includes factors such as desired --include features automatic
accuracy of calculation -patient data is captured and image allignment
stored correctly
-specific body measurements -include user friendly interface
need to be captured for easy navigation

designed the system hardware

and software components system need to be designed
improve privacy and security at
with proper authentication and
-selecting appropriate cameras patient data
authorization measures
-developing image processing -strict access controls and data
-only authorized personnel have
algorithms for analyzing encryption and regular backups
access to the patient information
captured image

system need to be tested and system need to be integrated

cost and maintenance
calibrated with health's center existing
patient information -cost of the system should be
-test subject within the budget of health
management
-comparing results center with minimal
-ensure the system will always
maintenance
up to date and secure

10