A Review of Object Detection Empowering Sports: Key Technologies, Application Scenarios, and Future Outlook

Object detection is now a cornerstone of 'Smart Sports,' yet the direct application of general-purpose models to the dynamic and often chaotic sports environment is fraught with challenges. This paper systematically reviews the core technologies of object detection in sports, including the adaptability and limitations of mainstream detectors (e.g., the YOLO series, Transformer-based models) in sports scenarios. It also examines the role of optimization strategies such as model pruning, quantization, and knowledge distillation in balancing performance and resource consumption, as well as specialized techniques for small object detection, motion blur processing, and occlusion robustness enhancement. Based on this, the paper provides an in-depth analysis of the diverse applications of object detection in professional sports training (e.g., motion capture and biomechanical analysis), competitive game analysis (e.g., tactical minimap reconstruction from match videos), intelligent officiating (e.g., foul recognition assistance), athlete performance evaluation, interactive sports broadcasting, and public fitness. Finally, the paper summarizes current challenges, including data bottlenecks, algorithm generalization, the complexity of multi-modal fusion, and the leap from perception to cognition. It also provides an outlook on future directions, including constructing sports-specific vision foundation models, deepening multi-modal intelligent fusion, enhancing dynamic scene understanding capabilities, and improving sports datasets and evaluation systems to promote the development of sports analytics toward intelligence, personalization, and accessibility.