Basketball Players Identification and Tracking using a Single Fixed Camera
Stephen Karungaru, Hiroki Tanioka, Kenji Matsuura, Kenji Terada · 2023
The application of data science in sports has recently gained significant importance. Consequently, tactics development, opponent team understanding, in-game decision-making, player vital statistics, etc. could all benefit from game data analysis. Given a video input, this study proposes employing consecutive intersection over union (IOU) based tracking and K-Means-based uniform colors detection methods to monitor players and identify teams in basketball. A retrained YOLOv7 algorithm is used to recognize the humans (players, referees, etc.), ball, and field markers. The Kalman filter is employed to tackle occlusion occurrence during tracking. Our findings indicate that this technology may be practical and affordable for both amateur and professional teams.