Research on Rock Climbing Motion Detection Algorithm Based on YOLOv8s-Pose

Professional training and safety monitoring of rock climbing highly rely on movement norm recognition. Existing visual detection methods are susceptible to occlusion, light and complex pose interference in dynamic rock wall environments, making it difficult to capture subtle joint changes. In this paper, we propose a lightweight improvement scheme based on the YOLOv8s-Pose architecture for the fast-transitioning limb movements in rock climbing. By fusing timing modeling and pose rule constraint mechanism, the problem of continuous action misdetection is solved. The research constructs a dedicated dataset adapted to the characteristics of rock climbing, which provides an automated monitoring technology foundation for high-risk sports and promotes the landing of intelligent sports analysis system.