摘要
分割与跟踪一直是视频图像处理的一个热门问题。针对足球视频应用的需要,本文实现了一个快速的球员分割与跟踪算法。根据颜色分量差值的统计信息,实现了球员的自动分割;通过归一化颜色直方图的相关性比较,辨识球员所属的球队;利用视频序列中球员的上下文特征,结合模板匹配方法,实现了球员的跟踪。实验结果表明,本文提出的跟踪算法很好地解决了不同球队队员的遮挡问题,并能处理多人遮挡情况,在球员形态变化较大时也能稳定地跟踪。
Segmenting and tracking are a hot problem in video processing. For the application of soccer video, this paper presents a method to implement quick segmentation and tracking of players from a soccer video. Players are segmented by the use of the statistics of color components differences; Correlation of two normalized color histograms is used to identify the team that a player belongs to; To track the players, this method has used the context features and correlated-template matching. Experiment results show that this method deals well with occlusion of two players belonging to different teams, and can track when multiple players are in occlusion. The method can track robustly when the players?shape changes greatly.
出处
《系统仿真学报》
CAS
CSCD
2001年第S2期90-93,共4页
Journal of System Simulation
基金
中科院计算所青年创新基金
国家自然科学基金项目(69673020)
国家重点基础研究发展规划973项目(G1998030607)资助。
关键词
分割
跟踪
上下文特征
模板匹配
相关性
segmentation
tracking
context features
template matching
correlation.