基于图像健美操动作精度监测研究.docVIP

基于图像健美操动作精度监测研究 　　摘 要： 针对目前大多数健美操动作精度监测都是通过对健美操动作进行三维重建，并进行健美操动作特征提取，建立健美操动作最优分类决策函数的基础上完成的，难以保证对健美操动作的所有特征进行提取，降低了健美操动作精度监测的准确性。为此，提出一种基于图像的健美操动作精度监测新方法，通过将健美操动作从背景中抠出来，利用Kinect深度图像采集方法对图像进行预处理，然后提出利用HOG3D对健美操动作特征进行提取，并分析提取过程。在此基础上，将健美操动作精度监测问题转化为视频内容分类问题，通过概率统计中的Adaboost方法对健美操动作精度进行识别，利用概率函数将健美操动作的姿势状态联系起来形成动作序列，并通过遍历函数取其最大值作为健美操动作的分类标准，选用逐层等比方法对健美操动作精度进行监测，从而实现健美操动作精度的监测。实验结果证明，所提方法能够有效地提高健美操动作轨迹监测的精度，降低健美操动作精度监测的能耗，具有良好的使用价值。 　　关键词： 图像； 健美操动作； 概率函数； 运动序列； 特征参数； 精度监测 　　中图分类号： TN911.73?34； TP391 文献标识码： A 文章编号： 1004?373X（2018）07?0075?05 　　Research on image?based movement accuracy monitoring of aerobics 　　FENG Ting 　　（Zhengzhou University of Aeronautics， Zhengzhou 450000， China） 　　Abstract： Most of the action precision monitoring of aerobics is realized by three?dimensional reconstruction of aerobics movement. The characteristic extraction of aerobics movement was performed on the basis of the optimal classification decision function， which is difficult to guarantee all the characteristics extraction of aerobics movement， and can reduce the movement monitoring accuracy of aerobics. An image?based new method to monitor the movement accuracy of aerobics is proposed to extract the aerobics movement in background. The Kinect depth image acquisition method is used to preprocess the image， and then the HOG3D is used to extract the action characteristic of aerobics and analyze the extraction process. On this basis， the problem of aerobics movement precision monitoring is transformed into video content classification problem. The Adaboost method in probability statistics is used to identify the movement accuracy of aerobics. The probability function is adopted to link the posture states of aerobics movement to form the movement sequences， and its maximum value taken by the traversal function is deemed as the classification standard of aerobics movement. The layer?by?layer equal ratio method is used to monitor the movement accuracy of aerobics， so as to realize the m