基于三维图像的运动员起跑动作误差预测仿真

摘要

田径比赛运动支撑期间,起跑动作的准确性可使运动员在最短时间内获得最佳速度,直接影响比赛成绩,因此需要对运动员起跑动作进行误差预测,当前方法预测起跑动作时,难以对运动员起跑动作关节点进行准确跟踪,降低了运动轨迹预测的精度.提出一种基于三维图像的运动员起跑动作误差预测方法.上述方法选用星型骨架组成结构描述运动员起跑动作三维运动模式,采用ISOMAP非线性降维方法计算获取起跑动作图像三维子空间的运动状态投影,将起跑动作三维数据投影至非线性低维子空间中,识别出运动员起跑动作状态的内在结构后研究整个起跑动作的各个关节点,利用Mean-Shift搜索算法确定运动员起跑动作各个关节点位置,并通过卡尔曼滤波器算法进行运动员起跑动作误差预测,确定运动员起跑过程中的动作轨迹.仿真结果表明,所提方法可有效提升运动员起跑动作轨迹预测精度,且预测效率较高.%This paper proposes a method for error prediction of starting action of athlete based on three-dimensional (3D) motion pattern.Firstly,the paper selected structure and composition of star skeleton to describe the 3D motion pattern and used ISOMAP nonlinear dimensionality reduction to calculate and acquire projection of athletic state of 3D subspace of starting action image,then mapped 3D data of the action into nonlinearity subspace with low dimension.Moreover,the inner structure of the action state was recognized and then every articulation point of all starting action was studied.The paper also used Mean-Shift search algorithm to confirm location of the articulation point,and on that basis,carried out error prediction for the starting action via algorithm of Kalman filter.Finally,we confirmed the motion locus during process of the starting.Simulation results show that the method can improve precision of motion locus of the starting action effectively and has high prediction efficiency.