This assignment introduces the student to developing programs that implement computer vision and image analysis methods in Matlab, which will be tested on given images and video sequences. Students are to develop Matlab programs that implement computer vision and image analysis methods to solve a given task (details will be given here in due time). Students are to show their knowledge of analysing images in general and working with object detection methods in particular to find foreground objects in scenes of different complexity. Students are to actively demonstrate their understanding and skills of feature extraction methods in the context of object detection and recognition. Students are to show their knowledge of image alignment using computer vision and machine learning methods..
The context for this assignment is the analysis of images and the recognition of objects in images from Robocup Soccer competitions. Your task is to wite Matlab code that automatically finds the location of the ball, the field lines, the goal and the robots by means of computer vision algorithms. Different solutions are possible, which means you are free to use any computer vision method (or combination of methods) that you want, All code must be in Matlab. If you are using code/functions/libraries outside those provided by Matlab’s toolboxes, you must reference your source(s) in the submission comments and in the code (via comments). By submitting the assignment, you declare that all code is your own unless referenced otherwise.
In the assignment, we will use images from 3 ‘levels’ of difficulty:
Images where the lines, ball and goal are clearly visible. (The lines may at times be blocked by the ball or robots.)
Images where the lines, ball and goal are partially occluded. That is, you might only see 3/4 of the ball or one post and the crossbar of the goal.
An image sequence where the ball and the robots move, thus creating a dynamic scene where occlusion may happen at some points in time, while the ball and goal may be completely visible at other time.
The training images below can be used during the development phase. They follow the same difficulty levels as mentioned above.
Image set for Level 1 and 2 [ZIP download ]
Video for Level 3 [MP4 download ] – Please download (rather than play), 3.1MB
You are to write Matlab code, using good software engineering skills, that will automatically analyse the images and display the location of
the field lines (highlight by overlay),
the goal (posts and crossbar – highlight by overlay),
the robots (highlight bounding box), and
the ball (highlight by circle or bounding box)
in the images and video.
Please put all your Matlab files into a single ZIP file and upload this ZIP file to Canvas.
I am looking forward to your great Matlab files.