王皓蜀



摘 要: 传统Mean Shift算法在运动目标运动速度过快以及被遮挡的情况下,算法的跟踪效果较差。因此,提出基于改进Mean Shift算法的网球运动视频目标跟踪方法,分析Mean Shift算法进行网球运动视频目标跟踪的过程以及存在的弊端。采用最小二乘法对Mean Shift算法进行改进,利用最小二乘法预测网球运动视频目标位置,在该位置上实施迭代跟踪,再用Mean Shift算法得到目标最终跟踪位置,解决目标运动速度过快以及遮挡问题的干扰,减小各帧检索时矢量同收敛点的距离,提高跟踪效率。实验结果说明,所提方法具有较高的跟踪效果和跟踪效率。
关键词: Mean Shift算法; 网球运动; 视频目标; 跟踪研究
中图分类号: TN911.73?34; TP391 文献标识码: A 文章编号: 1004?373X(2017)13?0073?04
Abstract: Since the tracking effect of the traditional Mean Shift algorithm is poor in the situations that the speed of the moving object is fast, and the moving object is blocked out, an improved Mean Shift algorithm based video object tracking method of tennis sports is proposed. The process and shortcomings of the Mean Shift algorithm to perform the video object tracking of the tennis sports are analyzed. The least square method is used to improve the Mean Shift algorithm, and predict the video object location of tennis sports. The iterative tracking is carried out for this position. The Mean Shift algorithm is adopted to get the final target tracking position, solve the problems of fast target movement speed and interference shielding, reduce the distance between the vector and convergence point when each frame is retrieved, and improve the tracking efficiency. The experimental results show that the method has high tracking performance and tracking efficiency.
Keywords: Mean Shift algorithm; tennis sports; video object; tracking research
0 引 言
随着计算机技术的快速发展,计算机视觉技术在人们的生产和生活中的应用价值也逐渐提升。视频目标跟踪是计算机视觉领域分析的热点,其广泛应用在各项体育运动训练和比赛过程中,对运动目标进行准确分类和跟踪,对提高网球运动的质量具有重要意义[1]。但是受到网球运动目标自身因素以及外部环境因素的干扰,使得网球运动视频目标跟踪算法的性能降低。特别是传统均值偏移跟踪算法(Mean Shift跟踪算法)在网球运动视频目标原始位置实施迭代的收敛过程中,在运动目标运动速度过快以及被遮挡的情况下,算法的跟踪效果较差[2]。因此,提出基于改进Mean Shift算法的网球运动视频目标跟踪算法,提高目标跟踪的效率和精度。
1 Mean Shift算法……p>