Design of Target Detection and Pose Analysis Algorithms for Basketball Free Throw Accuracy
Free throws are a crucial scoring method in basketball games and often have a decisive impact on the outcome of a match.In recent years,methods using vision or sensors for free throw analysis have become time-consuming and labor-intensive,making them insufficient to meet the demands of competitive sports.This study first employs the YOLOv5 network algorithm for the detection and tracking of target athletes,introducing an attention mechanism to enhance the algorithm.Subsequently,it utilizes the OpenPose network for human pose estimation after correcting misidentified key points.The results indicate that the proposed method stabilizes by the 29th epoch in loss curve comparisons,maintaining a loss of 0.352.In basketball video predictions on the ImageNet data,the method achieves an accuracy consistently above 91%,demonstrating high predictive accuracy for athletes'free throw success rates.This approach can be widely applied to various video datasets,providing a new reference for scientific and intelligent training in competitive sports.
human poserecognitiontarget detectionbasketballYOLOv5 Network
NETL
NSTL
万方数据
