复杂背景下的篮球识别

Basketball Recognition in Complex Backgrounds

针对篮球比赛视频分析应用,提出了一种多层次的篮球对象识别方法．利用篮球颜色一致性较好的特 点,通过彩色训练提取篮球颜色的统计范围,根据该统计范围得到图像中可能的篮球区域;根据篮球轮廓为圆的 特点,采取方向滤波器和Hough变换相结合的方法提取可能轮廓区域中的圆形轮廓,得到候选篮球;最后,利用 篮球典型的纹理特征,采用模板匹配的方法对候选篮球进行确认．实验结果证明了该算法的有效性,且多层次 算法可推广应用于其他球类对象识别．

Semantic object recognition is one of key issues in intelligent image understanding and so on. In this paper, we design algorithms to recognize basketballs in complex backgrounds. First, color feature is extracted to determine the possible regions of basketball. Then steerable filters are combined with modified hough transform to extract the circle contour, by which candidate basketballs are extracted. Finally, the candidate basketballs are confirmed using template matching by the texture image of basketbal...