Line detection and tracking in video recordings of rugby games : a thesis presented in partial fulfilment of the requirements for the degree of Master of Science in Computer Science at Massey University, Palmerston North, New Zealand

Thumbnail Image
Open Access Location
Journal Title
Journal ISSN
Volume Title
Massey University
The Author
Video analysis has long been used in sports analysis. More and more coaches and instructors choose to use computer-based video analysis systems in decision-making and player training, which makes computer-aided sports analysis a fast growing industry. AnalySports Ltd is a New Zealand based sports analysis company providing performance analysis for sports, especially for team games such as rugby. Currently AnalySports uses human coders to manually track the activities and position of the ball carrier. This thesis is part of a player-tracking project. The overall aim of the wider project is to build a cost-effective system to semi-automatically track individual players' positions in video recordings of team sport games. To obtain the information on movement and tactics of the team as a whole, it is necessary to identify the position of each player on the field at every point of time during the game. To perform the tracking, wide-angle video recordings from rugby games are used as input data. As the camera is moving, it is necessary to find the mapping between the pixels in the image and positions in the rugby field for every video frame. To make the size of the task realistic for a one-year masters project, the work presented in this thesis focuses on finding formulae and parameters for this conversion. Analysis of the position data for player performance and game tactics is outside the scope of this thesis. The conversion between image coordinates and field positions can be established by- identifying the field characteristics. The positions of players in the field then can be calculated using this conversion once they are identified on the video frame. Based on single video frames, algorithms have been developed to detect and identify field characteristics (lines) in the frames of the video recordings. Using the identified field characteristics as reference, a transform matrix was calculated to convert pixels in the image to positions in the rugby field. For a sequence of video frames, algorithms have been developed to track the identified reference lines in order to save time and human power. These tasks were complicated by the zooming, tilting, and panning movements of the camera and therefore a potential of loss of reference lines, the noise in the data caused by field properties such as advertisement, the varying light conditions, the movements of the sun. or the shadows of the stadium roof. An application was developed to perform the developed algorithms. The testing shows that for about seventy percent of the video clips investigated, lines can be recognised and tracked. That means the application can be used to find the conversion for the majority of the video clips. Based on the testing performed, further development based on this project could be a refinement of the image recognition parameters, efficiency improvements and the development of the actual player recognition. This project was supported in part by a grant from the (New Zealand) Foundation for Research, Science, and Technology (FRST) and the Technology for Industry Fellowships (TIF) programme (contract number: ANLY0201).
Optical pattern recognition, Digital video, Rugby football, Data processing