Tracking and detection in broadcast soccer video based on data fusion and YOLO model
Keywords:
Soccer video analysis, Object detection, Multi-object tracking, YOLOv8, Data fusion.Abstract
The broadcast soccer game video analysis system operates as a vital system which enables users to assess player performance and study team strategic methods. The process of reliable ball tracking during broadcast events faces multiple obstacles because the ball moves quickly while being small and because it often gets hidden by players and because the camera system continuously pans and zooms. The combination of these elements results in system failures which produce both incorrect tracking results and broken object paths and incorrect identification of tracked objects.
This paper presents a data fusion-based framework for joint detection and tracking of players and the ball in broadcast soccer video. The system uses YOLOv8 deep learning model for real-time object detection which then uses SORT algorithm to track multiple objects. SORT uses Kalman filter-based motion prediction together with IoU-based Hungarian association to preserve object identity consistency between different frames.
The system needs to perform preprocessing to remove non-play segments which contain replays and sudden scene changes for achieving better temporal stability. The system includes additional modules which perform camera motion compensation and homography-based field coordinate normalization and jersey color-based team classification and nearest-player assignment for ball possession estimation and speed and distance calculation for player performance analysis.
The researchers conducted experimental tests on Full HD broadcast videos which produced excellent detection results with precision at 0.96 and recall at 0.94 and F1-score at 0.95. The visual results show that the system identifies people correctly throughout all tests which used typical crowd numbers under conditions that mimic actual broadcasting scenarios. The proposed framework delivers organized spatiotemporal data which allows users to perform tactical analysis and performance assessment and operate automated sports analytics systems
Downloads
References
Reference
1. Rangasamy, K., As’ari, M. A., Rahmad, N. A., Ghazali, N. F., & Ismail, S. (2020). Deep learning in sport video analysis: A review. TELKOMNIKA, 18(4), 1926–1933. Doi:10.12928/TELKOMNIKA.v18i4.14730
2. Rahimian, P., & Toka, L. (2022). Optical tracking in team sports: A survey on player and ball tracking methods in soccer and other team sports. Journal of Quantitative Analysis in Sports, 18(1), 35–57. Doi:10.1515/jqas-2020-0088
3. Zhao, Z., Chai, W., Hao, S., Hu, W., Wang, G., Cao, S., Song, M., Hwang, J. N., & Wang, G. (2023). A survey of deep learning in sports applications: Perception, comprehension, and decision. arXiv. Doi:10.48550/arXiv.2307.03353
4. Bochkovskiy, A., Wang, C. Y., & Liao, H. Y. M. (2020). YOLOv4: Optimal speed and accuracy of object detection. arXiv. Doi:10.48550/arXiv.2004.10934
5. Bewley, A., Ott, L., Ramos, F., & Upcroft, B. (2016). Simple online and realtime tracking. In Proceedings of the IEEE International Conference on Image Processing (ICIP) (pp. 3464–3468). IEEE. Doi:10.1109/ICIP.2016.7533003
6. Yang, Y., & Li, D. (2017). Robust player detection and tracking in broadcast soccer video based on enhanced particle filter. Journal of Visual Communication and Image Representation, 46, 152–163. DOI: 10.1016/j.jvcir.2017.03.008
7. Wojke, N., Bewley, A., & Paulus, D. (2017). Simple online and realtime tracking with a deep association metric. arXiv. DOI: 10.48550/arXiv.1703.07402
8. Zhang, Y., Wang, C., Wang, X., Zeng, W., & Liu, W. (2021). FairMOT: On the fairness of detection and re-identification in multiple object tracking. arXiv. DOI: 10.48550/arXiv.2004.01888
9. Zhang, Y., Sun, P., Jiang, Y., Yu, D., Weng, F., Yuan, Z., Luo, P., Liu, W., & Wang, X. (2022). ByteTrack: Multi-object tracking by associating every detection box. arXiv. DOI: 10.48550/arXiv.2110.06864
10. Penn, M. J., Donnelly, C. A., & Bhatt, S. (2025). Continuous football player tracking from discrete broadcast data. Royal Society Open Science, 12, 251175. DOI: 10.1098/rsos.251175
11. Cioppa, A., Giancola, S., Deliege, A., Kang, L., Zhou, X., Cheng, Z., Ghanem, B., & Van Droogenbroeck, M. (2022). SoccerNet-Tracking: Multiple object tracking dataset and benchmark in soccer videos. CVPR Workshops. DOI: 10.48550/arXiv.2204.01888
12. Zhang, R., Wu, L., Yang, Y., Wu, W., Chen, Y., & Xu, M. (2020). Multi-camera multi-player tracking with deep player identification in sports video. Pattern Recognition, 108, 107260. DOI: 10.1016/j.patcog.2020.107260
13. Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. doi:10.1109/CVPR.2016.91
14. Kalman, R. E. (1960). A new approach to linear filtering and prediction problems. Journal of Basic Engineering.
15. Lucas, B. D., & Kanade, T. (1981). An iterative image registration technique. doi:10.1109/ICDAR.1981.4766994
16. Hartley, R., & Zisserman, A. (2004). Multiple view geometry in computer vision. Cambridge University Press.
17. Little, J. J., & Boyd, J. E. (1998). Recognizing people by their gait. doi:10.1109/CVPR.1998.698664
18. MacQueen, J. (1967). Some methods for classification and analysis of multivariate observations.
19. Deza, M. M., & Deza, E. (2009). Encyclopedia of distances. Springer.
20. Bishop, C. M. (2006). Pattern recognition and machine learning. Springer.
21. Everingham, M., Van Gool, L., Williams, C. K. I., Winn, J., & Zisserman, A. (2010). The Pascal Visual Object Classes (VOC) challenge. International Journal of Computer Vision, 88(2), 303–338. DOI: 10.1007/s11263-009-0275-4
