Article: Basketball player tracking method based on multi-source data and attention mechanism Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.46 pp.24 - 46 Abstract: Complex occlusions and rapid movements in basketball make tracking difficult, but deep learning-based visual computing provides effective new solutions. This study proposes an object tracking method that integrates the YOLOv5 model with the simple online and realtime tracking (SORT) algorithm. To address the challenge of multi-source information fusion, a cross-modal transformer model was designed to achieve adaptive deep integration of visual and motion data. Experiments utilised the public SportsMOT dataset, featuring 240 HD clips from real games across diverse arenas, lighting, and tactics. Validation on datasets shows the algorithm achieved a recall of 0.97 and a precision of 97.2%, with the mean average precision improving by 15% over the baseline. The multiple object tracking accuracy and precision reached 98.1% and 96.2% respectively. The algorithm thus proves to be an efficient and accurate tracking solution, offering robust data for coaching analysis and strategy. Inderscience Publishers - linking academia, business and industry through research

Title: Basketball player tracking method based on multi-source data and attention mechanism

Authors: Yu Lei

Addresses: School of Physical Education, Huazhong University of Science and Technology, Wuhan, 430074, China

Abstract: Complex occlusions and rapid movements in basketball make tracking difficult, but deep learning-based visual computing provides effective new solutions. This study proposes an object tracking method that integrates the YOLOv5 model with the simple online and realtime tracking (SORT) algorithm. To address the challenge of multi-source information fusion, a cross-modal transformer model was designed to achieve adaptive deep integration of visual and motion data. Experiments utilised the public SportsMOT dataset, featuring 240 HD clips from real games across diverse arenas, lighting, and tactics. Validation on datasets shows the algorithm achieved a recall of 0.97 and a precision of 97.2%, with the mean average precision improving by 15% over the baseline. The multiple object tracking accuracy and precision reached 98.1% and 96.2% respectively. The algorithm thus proves to be an efficient and accurate tracking solution, offering robust data for coaching analysis and strategy.

Keywords: tracking method; basketball players; multi-source data; attention mechanism; DeepSORT algorithm.

DOI: 10.1504/IJICT.2026.153521

International Journal of Information and Communication Technology, 2026 Vol.27 No.46, pp.24 - 46

Received: 14 Oct 2025
Accepted: 16 Dec 2025
Published online: 12 May 2026 *

Title: Basketball player tracking method based on multi-source data and attention mechanism

Keep up-to-date