基于方向预估计的边界跟踪算法  

Boundary tracking algorithm based on direction estimation

作　　者：田军委[1] 尚雅层[1] 王洪喜[1] 程钢[2] 

机构地区：[1]西安工业大学机电工程学院,陕西西安710032 [2]西安北方光电科技防务有限公司,陕西西安710043

出　　处：《应用光学》2011年第3期441-445,共5页Journal of Applied Optics

基　　金：陕西省科技计划项目(2010K08-16);陕西省教育厅专项科研计划项目(2010JK600);西安工业大学校长基金项目(XAGDXJJ0904)

摘　　要：为了实现多目标轮廓高效跟踪,提出了基于方向预估计的边界跟踪算法。研究了多目标轮廓模型表达以及轮廓初始点选取方法,为多轮廓跟踪提供模型依据。定义了4方向操作模板和方向矩阵实现轮廓走向的估计,通过定义权重矩阵避免轮廓跟踪算法进入搜索陷阱,提高算法的鲁棒性。对多组头部MR图像进行实验,结果表明,在同等边界数据情况下,新算法跟踪效率较变窗爬虫跟踪算法和传统跟踪算法平均提高约21%,鲁棒性明显增强。In order to obtain multi-boundary of digital image quickly,a new boundary tracking algorithm based on searching direction estimation is proposed.The model of multi-boundary and the initial feature point method were studied.Four direction operation templates and a searching direction matrix were defined to estimate searching direction of next step.For enhancing the robustness of tracking algorithm,a weight matrix was defined to keep tracking algorithm out of searching trap.Tests were performed on MR images,and the results show that the tracking efficiency of new algorithm is enhanced by 21% compared to variable window reptile method or traditional tracking algorithm.According to the test results,the robustness of the new tracking algorithm is improved obviously.

关 键 词：边界跟踪 方向预估计 多边界 鲁棒性 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]