现代电子技术2025，Vol.48Issue(18)：119-124,6.DOI:10.16652/j.issn.1004-373x.2025.18.019
融合局部感知增强的投篮上肢动作分解与高精度识别算法
Shooting upper limb motion decomposition and high-precision recognition algorithm integrating local perception enhancement
摘要
Abstract
In allusion to the problems of joint occlusion and large receptive field existing in traditional motion recognition algorithms when processing basketball training images,a high-precision upper limb motion decomposition recognition model integrating local perception enhancement is proposed.The model can extract the global structural features of the human body by means of lightweight high-resolution network(HRNet)branches,and combine with a two-stream pyramid module to enhance local joint perception ability.In the space flow,the cavity convolution is used to expand the hand receptive field,and in the time flow,the optical flow is used to capture the relative motion of the golfer,significantly improving the hand positioning accuracy in the occluded scene.At the same time,the spatiotemporal decomposition module is also designed in the model to analyze the spatial and temporal characteristics,constrain the network focus and learn the motion consistency between frames,and making the data processing process more robust.The adaptive weighting elements are introduced into the model to output high-precision joint point coordinates.The experimental testing results show that the mAP@0.5 of the proposed model can reach 86.9%,the F1 value of action classification is 95.3%,which is better than the current mainstream algorithm.and the key frame detection rate is 91.1%,the frame rate can reach 47 f/s,and the real-time performance is good,which fully proves that the model can provide reliable technical means for basketball training and intelligent analysis of events.关键词
HRNet模型/双流金字塔/图卷积网络/时间卷积网络/人体动作识别/图像分析Key words
HRNet model/two-stream pyramid/graph convolutional network/time convolutional network/human motion recognition/image analysis分类
信息技术与安全科学引用本文复制引用
任宇飞,刘海林..融合局部感知增强的投篮上肢动作分解与高精度识别算法[J].现代电子技术,2025,48(18):119-124,6.基金项目
西安2025年度社会科学规划基金(25TY05) （25TY05）
