跳远运动支撑期间下肢动作轨迹预测方法的研究

跳远运动支撑期间,下肢动作的准确性可使运动员最大程度避免受伤,且决定着运动员跳远成绩,因此需要对支撑期间下肢动作轨迹进行预测,以达到改善动作准确性的目的。传统方法预测动作轨迹时,一般采用光流估计法或变分方法;但这两种方法较为繁琐,且预测误差较大。为此,提出一种新的跳远运动支撑期间下肢动作轨迹预测方法,将跳远运动简化成运动员骨架的运动,获取三维人体骨架模型,通过未标定的多幅跳远运动支撑期间下肢动作图像对摄像机内参数进行确定,完成摄像机的自标定。将跳远运动支撑期间下肢动作的左右膝关节、左踝关节点作为关键节点进行分析,求出首帧图像对应的三维运动骨架特征点的坐标;在此基础上,继续求出各后续帧中三维人体运动骨架特征点的坐标,从而实现跳远运动支撑期间下肢动作轨迹预测。实验结果表明,该方法具有很高的轨迹预测精度。

The long-jump support during the motion of lower limb three-dimensional skeleton reconstruction

During the long-jump support, the accuracy of the lower limb movements can make athletes avoid injury to the greatest extent, and determines the jump performance athletes, so need to forecast the support during the period of lower limb movement trajectory, in order to achieve the purpose of improving action accuracy.Tradition method to predict motion trajectory, using the optical flow estimation method and variational method, but the two methods is relatively complicated, and the prediction error is bigger.For this, put forward a new kind of the long-jump support during the period of lower limb movement trajectory prediction method, the long-jump simplified into the movement of the skeleton athletes, obtaining 3 d human body skeleton model, through without calibration of the long-jump more support during the period of lower limb movement image to determine the camera intrinsic parameters, complete camera self-calibration.Will lower limb movements during the period of the long-jump support around knee and left ankle point as a key node is analyzed, and the first frame image corresponding to the three dimensional motion skeleton of the coordinates of the feature points, on this basis, continue to work out the 3 d human body skeleton in subsequent frames of the coordinates of the feature points, so as to realize the long-jump support during the period of lower limb movement track prediction.The experimental results show that the method has a high trajectory prediction accuracy.