基于帧间差分的足球球员检测算法  

Player detection algorithm based on consecutive frames in soccer video

作　　者：崔国栋[1] 柴林燕[2] 于明[1] 

机构地区：[1]河北工业大学计算机科学与软件学院,天津300130 [2]中国医学科学院生物医学工程研究所,天津300192

出　　处：《计算机工程与设计》2010年第7期1536-1539,共4页Computer Engineering and Design

摘　　要：针对帧间差分法不能很好应用于变化场景的缺陷,提出了一种改进的帧间差分法,通过边缘提取、区域补偿等操作将该方法应用到足球视频中,对球员进行有效的检测。该方法首先利用颜色空间统计信息反投影技术得到运动场地,利用球场的主色对球场边缘进行补色,区域补偿,得到固定场景。利用帧间差分法检测球员,最后通过对球员候选区域的颜色信息与模板直方图匹配,达到对球员的队属辨别的目的。实验结果表明,该方法检测效果很好且具鲁棒性。Aimed at the unsatisfactory result of the consecutive frames method for detecting moving objects in changing scene,a novel soccer player detection algorithm based on improved consecutive frames is proposed.Firstly,the method of dominate color of playfield back projection is used to filter the auditorium to get the playfield,then the color of playfield is extracted to fill the region outside,the fixed background is gotten,the consecutive frames method is applied to detect players.Finally,correlation of two normalized color histograms is employed to identify the team that a player belongs to.Experimental results show that the proposed algorithm is effective and robust.

关 键 词：球员检测 帧间差分 颜色直方图 反投影 足球视频 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]