Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit

COMP3527

Computer Vision

2022

Question 1

The Durham Apple Growers Company is considering updating its fruit packing facility to automatically grade apples based on their visual appearance.  They propose using multiple colour cameras installed above the conveyor belts that carry incoming apples from their delivery point to be packaged.

The system is required to be able to automatically determine the shape, size and surface colour characteristics of each apple.

Please present a design for such a system, to advise the company on various aspects of this task, that includes:

(a) a high-level design of how the 3D size of an apple can be estimated using images from one or more cameras.     [12 Marks]

(b) a high-level design of how an apple can be classified into one of a set of pre-defined colour categories using colour information from a single camera view.                         [8 Marks]

(c) an outline of an approach for measuring how well the exterior shape of the apple conforms to a circle.           [5 Marks]

Question 2

An autonomous driving company is considering adding a range of advanced com- puter vision algorithms on-board their vehicles.

The new algorithms are required to automatically track pedestrians and vehicles that appear in the images from various cameras on-board the vehicle, with the aim of monitoring their trajectory relative to the motion of the vehicle.

Please present a design for such a system, to advise the company on various aspects of this task, that includes:

(a) a high-level design fully detailing how pedestrians and vehicles can initially be detected within the images prior to tracking. [11 Marks]

(b) a high-level design detailing how detected objects can be tracked within a single camera view. [9 Marks]

(c) an analysis of potential sources of tracking failure and how these could potentially be overcome to re-acquire a lost tracking target.     [5 Marks]

Question 3

A company producing autonomous lawn-mower robots is considering various op- tions for the processing of video images from the on-board cameras in order to provide (1) autonomous navigation on lawn environments and (2) obstacle avoidance for safety.

Each robot is equipped with a low-cost sensing payload comprising of an omni- directional camera providing a 360◦  colour image of the environment and also a joint Global Positioning System (GPS) / Inertial Measurement Unit (IMU).

(a)  Please present a design for the sensing component of the robot autonomous

navigation system  based on the current sensing  payload,  that  provides details of both:

i.  how the position and trajectory of the robot within the environment, relative to other objects, can be determined,    [10 Marks]

ii.  how accurate position estimates and distances to scene objects can be obtained from this single camera sensing payload.       [3 Marks]

(b)  Please present a detailed design for an obstacle avoidance system, based on the current sensing payload of the robot, that details how generic obstacles on the lawn could be detected in a way that allows the robot to navigate around them. [12 Marks]

Question 4

The Bank of Durham is considering the use of a new computer vision system to automatically verify UK bank notes paid into its branches.

This will be performed using high-speed cameras,  placed inside the banknote counting machines in each branch, that will image each and every banknote as it is counted.

The computer vision system will be required to automatically (1) determine the denomination of the banknote and (2) detect any damage or defacement of the banknote.

(a)  Please present a design for a computer vision approach to determine the

denomination of the banknote that includes:

i. a high-level design detailing key algorithmic approaches to be used, [6 Marks]

ii. any data set requirements and how these should be utilised, [4 Marks]

iii.  how operational performance could be monitored. [2 Marks]

(b)  Please present a design for a computer vision approach to detect any dam-

age or defacement of the banknote that includes:

i. a high-level design detailing key algorithmic approaches to be used, [6 Marks]

ii. any data set requirements and how these should be utilised, [4 Marks]

iii.  how counterfeit banknotes could potentially be detected. [3 Marks]