Assignment 1: Image Processing and Computer
Vision (20%)
COMP 4180 Intelligent Mobile Robotics
FALL TERM 2020
Release Date: 17 September 2020
Due Date: 13th October 2020, 4 PM (UTC -6 Central Time)
Content
In this assignment, you will use the OpenCV 3.3x, a standard vision library to develop a vision
module to support capturing and manipulation of images and videos. This vision module will be
carried over from assignment to assignment as I mentioned in class. There could be several
solutions to these open-ended problems. So general assessment will not only be based on the
outcomes but also on functionality, efficiency, and style, e.g. how elegant your code is.
Object and Feature Tracking Using OpenCV Library
Implement a feature detector that can reliably detect the feature point and recognise objects such
as obstacle, region of soccer field, goal, etc. in RoboCup Humanoid League our Simurosot
competition. Motion Analysis and Object Tracking could be useful too.
1
Figure 1: SimuroSot Robo Challenge Obstacle Avoidance
Figure 2: Expected outputs.
Tasks
1. Download and install OpenCV 3.3.x in your Linux machine (Ubuntu 16.04).
2. Download the sample POV images here.
3. Preprocess the input image from the streaming camera, converting it to different colour
space (RGB, HSV, grayscale), blurring, and detecting edges (each of these is a call to an
OpenCV method – such as Canny() for canny edge detector). Documentation on the
methods can be found on the OpenCV site: http://docs.opencv.org/.
2
4. [3%] Goal Detection: Write a subroutine called findGoal find the goal from the soccer field
from the sample POV images, as shown in Figure 2. Hint: extract the green soccer field
from the background such as the floor and surrounding. For example: return (x,y,area)
5. [5%] Obstacle Detection: Write a subroutine called findObstacle to find the obstacles in the
soccer field as shown in Figure 1 and 2. This subroutine should return the position (in 2D
space) and size of the obstacle. You will need to use image processing methods such as
colour extraction, thresholding, shape recognition, and morphological operation. Hint: You
may use ApproxPolyDP to detect the contour of the obstacles.
6. [10%] Write another subroutine called drawField to detect lines segments in the soccer field
and draw coloured lines to differentiate various region of the field such as goal lines
(pink), touch line (blue), center circle (yellow), etc. to distinguish the output visually
as shown in Figure 2. You need to implement a geometric approach based on line segment
detector fast line detector, and/or corner detector. Hint: Hough Line Transform is too
expensive.
7. [2%] Write a 2-page report (11-point, Arial, single-space) to explain the implementation and
results based on task 4 to 6. Include some of image processing outputs as an image
pipeline in your report.
8. Make sure your codes are well commented in order to obtain full marks.
9. Machine learning approach is a plus!
Submission
This assignment shall be done individually.
You must submit all parts of the assignment before the due date and time. Create a zip or tgz
archive which includes all source code of your project. Your submission should extract into a
directory called _a1.
Write a README file to explain anything you feel is necessary or important about your submission.
This may include special features/bugs of your program. Describe what parts of the assignment
you implemented. It is in your interest to simplify the job of the marker.
3
Submit/upload the archive to me via UMLearn. You have to agree to the honesty
declaration on “Checklist” at UMLearn before the submission.