Winter Examination Period 2021 – January – Semester A
ECS709 Introduction to Computer Vision Duration: 3 hours
This is a 3-hour open-book exam, which which must be started within a 24-hour period.
You MUST submit your answers within 3 hours of the time that you started the exam.
Follow all instructions on the download page.
You can refer to textbooks, notes and online materials to facilitate your working, but normal
referencing and plagiarism rules apply, and you must cite any sources used.
You must upload a single PDF document containing your solutions. These can be typed or
handwritten, or a mix of the two. Multiple submissions are not permitted, so be sure that you check
your submission before uploading it.
Calculators are permitted in this examination.
Answer ALL questions.
You must adhere to the word limit specified in the questions. Failure to do so will lead to
those answers not being marked.
YOU MUST COMPLETE THE EXAM ON YOUR OWN, WITHOUT CONSULTING OTHERS
Examiners:
Professor Andrea Cavallaro, Dr Miles Hansard
© Queen Mary University of London, 2021
Page 2 ECS709 (2021)
Question 1
Consider a camera placed on a third-floor window of a residential building. The field of
view of the camera is fixed and covers a pedestrian crossing at 3pm on a sunny day.
(a) Describe each step of a computer vision pipeline that counts the number of
pedestrians crossing the street during a 30-minute interval. Discuss your choices
for this pipeline and their potential limitations.
[10 marks]
(b) Discuss what problems shadows cast on the pavement may cause for each step of
your computer vision pipeline. Discuss what color space and what method (or
methods) you would use to segment the shadows cast by the pedestrians on the
pavement.
[10 marks]
(a) Discuss what segmentation errors you expect will occur and how you could
objectively evaluate the quality of the shadow segmentation results. For the design
of the evaluation process, assume that you have no constraints on the available
budget, time and computational resources.
[10 marks]
(c) List at least one potential ethical issue you foresee for the computer vision
application described in your answer to Question 1(a) and discuss how this issue
could be addressed or mitigated.
[10 marks]
Answer Question 1 with at most 1300 words. A diagram or illustration is considered to be
equivalent to 100 words. An equation is considered to be equivalent to 10 words.
ECS709 (2021) Page 3
Turn Over
Question 2
Consider an agricultural drone equipped with a camera to monitor crop growth and help
increase crop production.
(a) Discuss what texture descriptors you would use to design a computer vision
pipeline to distinguish crops from buildings and roads, and to recognize different
crops. Motivate your choices.
[10 marks]
(b) Discuss how you would design a computer vision pipeline that generates a single
image by merging the images captured in 10 consecutive frames of the video taken
by the drone. Discuss each step of this pipeline and list the challenges for each
step.
[10 marks]
(c) Describe each steps of a computer vision pipeline to count the number of vehicles
seen by the camera during a mission of the drone. Compare your choices for this
pipeline with those you made for the computer vision pipeline you described in your
answer to Question 1(a).
[10 marks]
Answer Question 2 with at most 1000 words. A diagram or illustration is considered to be
equivalent to 100 words. An equation is considered to be equivalent to 10 words.
Page 4 ECS709 (2021)
Question 3
Consider an unmanned underwater vehicle (or underwater drone) for the visual inspection
and classification of coral banks in shallow waters.
(a) For navigation and for self-localisation, the robot needs to match (at different time
instants) portions of the underwater scene seen from different viewpoints and at
different scales. Describe each step of a computer vision pipeline to localise feature
points in each frame and to match these feature points across frames. Discuss the
choice of the feature point detector and of the feature point descriptor.
[10 marks]
(b) List and discuss the main challenges of underwater computer vision tasks
compared to computer vision tasks for images and videos taken in air.
[10 marks]
(c) Describe the design of a computer vision pipeline to reduce (or remove) the effect
of scattering and colour cast from the corals, as captured by the camera, in order to
restore their colours as they would appear outside the water. If you propose to use
a machine-learning based approach for this task, discuss how you would select the
training set. Assume that you have no constraints on the available budget, time and
computational resources.
[10 marks]
Answer Question 3 with at most 1000 words. A diagram or illustration is considered to be
equivalent to 100 words. An equation is considered to be equivalent to 10 words.
End of Paper

欢迎咨询51作业君