Sports and Surveillance Video Analysis and Annotation

Kaelon Lloyd


Supervised by Dave Marshall; Moderated by Paul L Rosin

There a good few potential projects in the area of Video Analysis available:

1. Sports Video: This work builds on an on going collaboration with the Welsh Rugby Union and Swansea University. Every game of Rugby Union (both International and Region/Club) is annotated by hand by expert sports coders employed by the Welsh Rugby Union. Key events like Tries, Kicks, Scrums, Lineouts, Tackles etc. are annotated using specialist software (www.sportstec.com/Products_Sportscode). This is outputted in XML format and is synchronised to accompanying video data.

2. Surveillance Video: An alternative source of data but with similar goals is that of detecting key events in city centre surveillance cameras. Video data from Cardiff Police is available and the project will work in collaboration with Cardiff Univeristy's Dental School (Alcohol and Violence Reasearch Group) and the School of Social Science's Police Science Unit.

The basic idea of for all the above projects is to utilise this rich source of data and employing machine learning and computer vision techniques develop tools to assist in the annotation of video. The data can be mined to:

* assist in the tedious hand labelling of video (the current norm) * recognise key events (e.g. Rucks, Lineouts in Rugby, Fights, potential flash points in Surveillance Video. * provide a more quantitative analysis of data between key events, e.g. distance between events * predict future events.

An interest in image/video processing, modern computer vision and machine learning techniques is essential for this project. A keen interest in sport and Rugby in particular is also a distinct advantage for certain flavours of the above projects.

Knowledge of OpenCV and/or MATLAB video processing also useful.

Initial Plan (03/02/2014) [Zip Archive]

Final Report (06/05/2014) [Zip Archive]

Publication Form