摘要
提出了一种基于模板和改进的gradient vector flow(GVF)分割方法。该方法先手工建立初始化模板,利用初始化模板和分割对象的周期性线性匹配,并基于Chamfer距离寻找最佳匹配模板;把该最优模板轮廓作为改进GVF的初始轮廓,再使用改进的GVF算法分割出对象。该算法仅需建立一次初始化模板,以后具有通用性,而且对于阴影和背景影响有较好的分割效果。对加利福尼亚大学步态数据库研究显示了该方法的有效性。
This paper proposed a new method.Firstly,based on initial templates established manually,obtained the template according to object segmented by linear rule between manual templates and object cycle.Then,used Chamfer distance to search optimal position and template,whose contour was initial snake points.Lastly,improved the algorithm of GVF,which was used to calculate fine contour.This paper established initial templates only once,and the method had better segmental results on shadow.It is displayed that the algorithm is effective to UCSD gait database.
出处
《计算机应用研究》
CSCD
北大核心
2012年第3期1116-1118,1126,共4页
Application Research of Computers
基金
国家"863"计划资助项目(2008AA12A218-51)
