基于主动轮廓模型的红外图像轮廓提取算法  

Improved contour extraction algorithm of infrared images based on active contour models

作　　者：董恩增[1] 冯倩[1] 于晓[1] 佟吉钢[1] 谷海清[1] 

机构地区：[1]天津理工大学自动化学院,天津300384

出　　处：《激光与红外》2017年第3期379-384,共6页Laser & Infrared

基　　金：国家自然科学基金项目(No.61502340);天津市高等学校科技发展基金计划项目(No.20120829);天津市应用基础与前沿技术研究计划项目基金项目(No.15JCYBJC51800)资助

摘　　要：针对主动轮廓模型存在的对初始轮廓位置敏感、凹性目标轮廓无法正确收敛等问题,本文将自适应边缘检测和主动轮廓模型相融合,提出一种改进的红外图像目标轮廓自动提取算法。首先,采用最大类间方差法计算红外图像边缘检测算法的自适应阈值,获取目标初次边缘,降低对初始轮廓位置的敏感性;对初次边缘分别进行横向与纵向填充,填充图像相与运算,对得到目标区域提取二次边缘,将其作为主动轮廓模型的初始轮廓,保证目标凹陷区域轮廓的有效收敛。最后,通过仿真分析验证了该方法能够实现红外目标轮廓的精确自动收敛。Because active contour models are sensitive to the position of initial contour and concavity target contour can＇t converge correctly,an improved algorithm for extracting target contour of infrared images automatically was proposed based on adaptive edge detection and active contour models. Firstly,for reducing sensitivity to the position of initial contour,the adaptive threshold was calculated by OTSU algorithm,and then the first edge would be extracted. Secondly,the first edge was filled horizontally and vertically. For getting the target area,filling images were computed by using AND operation. The second edge which was extracted from the target area was considered as the initial contour of active contour model. This could ensure the effective convergence of boundary concavities. The simulation analysis results show that the infrared target contour could be converged accurately and autonomously by using this method.

关 键 词：自适应边缘检测 最大类间方差法 图像填充 主动轮廓模型 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]