网球视频中球员的侦测与追踪

针对现有的球员侦测与追踪方法中颜色信息的提取容易受到目标与噪声之间的色差及光照变化的干扰,从而影响侦测准确度的问题,提出了一种球员侦测与追踪的新方法.首先运用较为可靠的边缘信息侦测网球球场的场线,重建场地信息.其次运用重建场地信息设计过滤器,过滤球场噪声; 运用场线信息,定义上下半场球员侦测的范围与两种不同尺寸的搜寻视窗.然后通过提取图像边缘密度信息,在搜寻窗内利用模板匹配方法估测球员位置.最后利用3种场地类型对本文方法进行了验证,结果表明本方法的平均正确率(92.3%)高于文献3]的方法(91.12%),且比文献3]更具有普适性.

Detection and tracking of tennis video players

The existing player detection and tracking methods from videos usually utilize color information, which will reduce the detection accuracy since color information extraction is easily affected by noises and illumination change. To solve the problem this paper proposes a new method for player detection and tracking. Firstly, the more reliable edge information is utilized to detect the field lines and reconstruct the court accordingly. Secondly, the reconstructed court information is used to filter out the noises outside the court. Next, we use the field line information to calculate the detection ranges and search window sizes of players in the upper half and the lower half courts. Finally, the image edge density information is extracted, and the template matching method is applied in the search window to estimate the player position. The results show that the average correct rate of this method(92.3%)is higher than that of the literature 3](91.12%), and good adaptability to the tennis courts of different materials is also proved.