Conference paper
Computer vision in netball
Proceedings of the 15th ANZIAM Mathsport , pp.70-78
Australasian Conference on Mathematics and Computers in Sport: Mathsport , 15th (Online, 09-Nov-2020 - 11-Nov-2020)
Australian and New Zealand Industrial and Applied Mathematics
2020
Abstract
Understanding the location of a player and the positional relationship to their teammates and opposition is fundamental in sports analysis. Acquiring this information is challenging and traditionally relied on subjective, labour intensive manual notation, or compiling video edits. Although wearable devices have made this process more efficient, they are still limited in their application as team analysts don’t receive location data of the opposition. Computer vision, a statistical modelling process of classifying or detecting an object of interest in an image or video presents a practical option to acquire location data of players in a match. Computer vision has been successfully implemented using the combination of multiple camera video to provide the model with optimum conditions for player detection. Due to the extensive cost and non-portable nature of this set-up, its application is generally limited to large TV broadcast installations. This research presents a preliminary study of the ability and practicality of performing a player location process using a single fixed camera computer vision application within netball. Netball presents as a challenge to computer vision due to the fast rate of play, and the erratic movement of players through a confined space. Results from this study show that using computer vision to define player location in netball is possible. The application of an Aggregated Channel Features (ACF) model on 5391 ground truth images of players was shown to provide sufficient data to detect players in match conditions, and derive a valid representation of player location on the court. This research outlines the process used and the difficulties in its implementation as part of a team analysis system. It also targets areas of model improvement for future research. The findings from this study highlight the potential strength of using computer vision in netball to objectively define player location and provide context to key performance indicators.
Details
- Title
- Computer vision in netball
- Authors
- Paul Smith (Author) - University of the Sunshine Coast, Queensland, Faculty of Science, Health, Education and EngineeringAnthony Bedford (Author) - University of the Sunshine Coast, Queensland, High Performance Sport - Legacy
- Contributors
- Ray Stefani (Editor)Adrian Schembri (Editor)
- Publication details
- Proceedings of the 15th ANZIAM Mathsport , pp.70-78
- Conference details
- Australasian Conference on Mathematics and Computers in Sport: Mathsport , 15th (Online, 09-Nov-2020 - 11-Nov-2020)
- Publisher
- Australian and New Zealand Industrial and Applied Mathematics
- Organisation Unit
- School of Health and Sport Sciences - Legacy; School of Health and Behavioural Sciences - Legacy; University of the Sunshine Coast, Queensland; Faculty of Science, Health, Education and Engineering; High Performance Sport - Legacy; School of Science, Technology and Engineering
- Language
- English
- Record Identifier
- 99488708602621
- Output Type
- Conference paper
Metrics
22 Record Views