多特征光流跟踪人体关节点的运动员姿态分析模型

Athlete 's Attitude Analysis Based on Multi-Characteristic Optical Flow Tracking Human Joint

为了提高运动员姿态分析的精确性,以量化运动员视频数据进行训练反馈,本文采用图像处理技术对运动员视频进行处理,提出了一种多特征光流跟踪人体关节点的运动员姿态分析模型。首先采用垂直积分投影、水平线扫描、索引查找表和长度比例约束相融合的多特征算法进行关节点提取,然后采用Lucas-Kanade光流跟踪算法对关节点的运动进行初步跟踪,最后采用比例正交投影模型将人体二维骨架模型变换到三维空间,获得人体三维姿态。模型仿真实验结果表明,本文所提出的动作识别及姿态分析系统可以很好地识别人体动作,从而进行运动分析,为体育运动训练提供可视化的帮助。

In order to improve the accuracy of the athlete's attitude analysis and to train the athlete's video data, this paper adopts the image processing technology to deal with the athlete's video, and proposes a multi-characteristic optical flow to track the athlete's attitude analysis model. Firstly, the multi-feature algorithm is used to extract the joint points, and then the Lucas-Kanade optical flow tracking algorithm is used to track the movement of the joint nodes. Finally, the proportional orthogonal Projection model transforms the human body two-dimensional skeleton model into three-dimensional space, obtains the human body three-dimensional posture. The results of model simulation show that the motion recognition and attitude analysis system proposed in this paper can well identify the movement of the human body, so as to provide the visual help for the sports training.