2025 SURF项目展示——AI篮球视频分析系统来了!

我爱免费

作者：微信文章


















AIAC

AI篮球视频分析

   系统来了！

2025 SURF项目展示











AIAC





你是否想过，AI能否实时了解比赛中球员的位置和队伍布局？我们的团队开发了一套基于YOLOv11的篮球视频分析系统，可自动处理比赛视频，为教练、解说和球迷提供全新视角。



系统已实现功能



聚焦科技与未来的核心主题，精准锚定科技创新驱动发展的核心逻辑。科技创新是突破发展瓶颈、开拓未来赛道的核心动力。

球员&裁判检测：精准识别场上每位球员和裁判

跟踪：使用DeepSORT追踪，尽量保持身份一致性，重建球员轨迹

队伍分类：通过颜色特征半人工精校自动高精度区分两队球员

骨架估计：集成RTMPose-S，预测球员关键关节点

投影可视化：将球员位置映射到篮球场平面，直观观察队伍布局

实验成果



在真实比赛视频中，系统检测准确率达到98%，队伍分类准确率95%，展示了稳定且高效的分析能力。生成的可视化视频和JSON数据可用于进一步研究或教学示范。

未来预期



自动识别球员动作，如投篮、运球

高级战术分析，辅助教练决策和比赛解说





图1：骨架模型



图2：队伍分类



图3：2D 投影可视化

学生：焦英博、苗一诺、张文婧、苏泓铭、黄皓洋

指导老师：周冕博士

Scroll up to read

AI-Powered Basketball Video Analysis!

Ever wondered if AI can understand player positions and team layouts in real time? Our team developed a YOLOv11-based basketball video analysis system, which automatically analyzes game videos, providing a new perspective for coaches, commentators, and fans.

Current Features:

Player & Referee Detection: Accurate identification of all players and referees on the court

Tracking: DeepSORT maintains consistent IDs, reconstructing player trajectories

Team Classification: Automatically and precisely distinguish the two teams of players through color feature semi-artificial fine-tuning.

Pose Estimation: RTMPose-S predicts key body joints for each player

2D Court Projection: Map player positions onto a basketball court plane for intuitive team layout visualisation

Results:
On real game videos, the system achieved 98% detection accuracy and 95% team classification accuracy, demonstrating stable and efficient performance. The visualised videos and structured JSON outputs support further research or teaching applications.

Future Plans:

Automatically recognise player actions such as shooting or dribbling

Advanced tactical analysis to support coaching decisions and game commentary