2020 Oct 16 CSCI435 / CSCI935 Spring 2020 Page 1 of 4
School of Computing and Information Technology
Student to complete:
Family name
Other names
Student number
Table number
CSCI435 / CSCI935
Computer Vision Algorithms and Systems
Wollongong and SWS Campus
Examination Paper
Spring Session 2020
Exam duration 2 hours
Items permitted by examiner Open Book
Aids supplied Nil
Directions to students 4 questions to be answered.
Each question is worth 10 marks.
This paper is worth 40% of the total marks for the subject
Answer each question on a separate page clearly
Convert the answers into one pdf file
Submit the pdf file
2020 Oct 16 CSCI435 / CSCI935 Spring 2020 Page 2 of 4
4 Questions – 10 marks each Total: 40 marks
Please answer the questions clearly and with as much detail as possible. Please start each answer on
a new page.
1) Quality of images includes sharpness and colour fidelity. Nowadays, most digital cameras are
single sensor-based.
a) Name the major components (excluding lens) in a single sensor based digital camera and
describe the function/purpose of each component.
b) List three components that would affect most significantly sharpness of the captured images
and explain how each of these three components affects the sharpness.
c) List three components that would affect most significantly colour fidelity of the captured
images and explain how each of these three components affects the colour fidelity
d) Most cameras nowadays provide a capturing model called HDR (high dynamic range) for a
situation where the range from the lowest illumination to the hightest illumination in the
scene at the time when the image is being taken is extremely high. Explain the common
approach to or method for achieving HDR imaging in most of the digital cameras.
e) If an image captured by a camera appears to be dark and you wish to see the details in the
dark areas, suggest a possible method to enhance the image and explain HOW the method
would work to make the details in the dark regions more visible.
2) You need to implement a system that can automatically detect car registration plates (yellow
background and black digits and letters) in an image and outline the plates with a bounding box
as shown below.
Figure 2.1: Detection of the car registration plate in an image and extraction of its bounding box
Operational conditions: frontal view, good lighting conditions, good quality camera and near
distance as shown in Figures 2.1 and 2.2 (more images to illustrate the conditions).
2020 Oct 16 CSCI435 / CSCI935 Spring 2020 Page 3 of 4
Figure 2.2 Sample images of car registration plates
a) Propose a solution to the problem. The solution takes a colour image as input and outputs
the bounding box of a registration plate, if any. Divide the solution into components and
describe the solution using a block diagram or flowchart. Explain the function, input and
output of each components.
b) For each component in the solution, choose suitable algorithms and briefly describe how the
algorithms works.
c) Describe how you would test your solution and measure its performance.
d) Discuss whether your algorithm would work in a raining or snowing day and in a night.
Explain why it works or why it does not work.
e) Discuss whether your solution would be able to extract multiple registration plates in a
single image if there are two or more cars in the image. Explain why it is or why it is not.
3) A stationary video camera is often installed on the ceiling of a store entry to count how many
people are entering and/or leaving the store. Figure 3 illustrates the camera setting:
Front & top views
Shopping trolley
Figure 3 Overhead people counter: in and out
a) How will you classify this problem with regards to computer vision problems you have
studied in the subject?
b) Based on the algorithms you have studied in the subject, design a solution (block diagram)
that counts people without distinguishing entering and leaving.
c) Design/select AND describe the algorithms for each block/component in your solution.
d) What are the possible factors that may affect the accuracy of your system?
e) How would you modify your solution or algorithms such that the system is able to count the
number of people leaving the store with shopping trolleys?
2020 Oct 16 CSCI435 / CSCI935 Spring 2020 Page 4 of 4
4) You are asked to design a computer vision system to classify the vehicles into three major types:
trucks, sedans and vans, by only analysing a single image taken by a camera installed in a
free-way exit gate for every vehicle passing through the gate. Following are some sample
images.
Figure 2. Samples images of trucks (1st row), vans (2nd row) and sedans (3rd row)
a) How will you classify this problem with regards to computer vision problems you have
studied in the class?
b) Based on the algorithms you have studied in the class, propose a solution that takes images
and output the category of the vehicle in the images. Describe the solution in a block
diagram and explain the function of each components of the solution.
c) Design/select AND describe the algorithms for each block/component in your system.
d) Based on your solution, which type of vehicles is likely to have a high classification
accuracy or likely to be easily classified, explain why this is the case in the context of your
solution?