体育训练过程中的运动视频分析与识别研究  

Study on analysis and recognition of moving video in physical training

作　　者：南秋红[1] 

机构地区：[1]黄河科技学院,河南郑州450003

出　　处：《现代电子技术》2017年第11期68-71,共4页Modern Electronics Technique

基　　金：河南省科技厅科技攻关项目(132102310462)

摘　　要：当前的体育训练过程中,训练员使用难度较小的视频重播与解析管理方式为运动员讲解动作要领,不够直观和科学,不能满足训练员对运动效果评估的需求。针对该问题,研究了体育训练过程中的运动视频分析与识别过程,采用基于粒子滤波预测的自适应阈值运动目标分离算法实现运动目标的自适应分离。通过粒子滤波技术跟踪运动员的运动,塑造运动模型,并依据运动模型预测后续运动视频帧内不同重要关节点的位置,完成后续运动视频帧的跟踪。采用条件随机场方法实现体育训练视频中的动作识别。实验结果说明该方法具有较高的动作识别率和较低的误分离率。The trainer replays the low difficulty action video and uses the analysis management way to explain the action essential for athletes in physical training,which is not intuitive and scientific enough,and can′t meet the demand of trainers for sports effect assessment. Aiming at the above problems,the motion video analysis and recognition process in sports training are studied,and the adaptive threshold moving target segmentation algorithm based on particle filtering prediction is employed to realize the adaptive segmentation of moving target. The movement of athlete is tracked with particle filtering technology to shape the motion model,and predict the locations of articulation-points with different importance in the subsequent moving video frame according to the motion model,so as to track the subsequent moving video frame. The condition random field method is adopted to realize the movement recognition in sports training video. The experimental results indicate that the method has high movement recognition rate and low false separation rate.

关 键 词：体育训练 运动视频 分析与识别方法 粒子滤波技术 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP391[自动化与计算机技术—计算机科学与技术]