基于BlazePose与深度相机的体育评价系统  

Sports Evaluation System Based on BlazePose and Depth Camera

作　　者：吴苏皖 徐荣青[1] 孔梅梅[1] WU Suwan;XU Rongqing;KONG Meimei(College of Electronic and Optical Engineering&College of Flexible Electronics(Future Technology),Nanjing University of Posts and Telecommunications,Nanjing 210000)

机构地区：[1]南京邮电大学电子与光学工程学院、柔性电子(未来技术)学院,南京210000

出　　处：《计算机与数字工程》2024年第8期2306-2311,共6页Computer & Digital Engineering

摘　　要：在目前的体育自动化测试和训练中,往往需要受试者穿戴多种传感器,影响了成绩的发挥,另一方面,仅依靠二维图像或深度相机的动作捕捉方法也难以全面、准确地描述人体运动姿态。论文针对上述问题,提出一种适用于常规体测项目的动作评价系统,首先将二维人体姿态估计模型BlazePose与深度相机结合,实现人体姿态关键点在三维空间中的准确定位,再通过卡尔曼滤波对关节点的轨迹进行平滑与估计,在此基础之上根据有限状态机的思想,以引体向上为例,完成了对多指标体测项目评价方法的设计,能够全面检测出各种犯规动作,实验表明整体正确检测率达到89%,处理速度达到20FPS,能够用于实际的辅助测试以及对个人的训练起到指导作用。In the current automated sports testing and training,subjects often need to wear a variety of sensors,which affects the performance.On the other hand,it is difficult to describe the human motion posture comprehensively and accurately only relying on two-dimensional images or depth cameras.Aiming at the above problems,this paper proposes a motion evaluation system suitable for conventional physical measurement projects.Firstly,the 2d human pose estimation model BlazePose is combined with the depth camera to achieve the accurate positioning of key points of human pose in 3d space,and then the track of key points is smoothed and estimated by Kalman filter.On this basis,according to the thought of finite state machine,pull-ups,for example,the evaluation method for multi-index cervix projects design is completed to fully detect all kinds of fouls,experiments show that the overall detection rate reaches 89%,correct processing speed up to 20 frames per second,which can be used in the actual test and play a guiding role for personal training.

关 键 词：人体姿态关键点 动作评价 体育测试 深度相机 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]