基于两方法博弈的马尔可夫随机场红外图像分割算法  

Infrared Image Segmentation Based on MRF Combined with Two-algorithm Game

作　　者：王坤[1] 张恺[1] 王力[1] 诸葛晶昌[1] 

机构地区：[1]中国民航大学,天津300300

出　　处：《红外技术》2015年第2期134-138,共5页Infrared Technology

基　　金：中国民航大学校级项目基金;编号:2010ky E07;2010QD02S;国家自然科学基金;编号:U1333111;天津市自然基金;编号:12JCQNJC00600;中国民航机务维修科研基地资助

摘　　要：电路板红外图像芯片提取是电路板红外故障检测系统中的重要环节,已成为红外图像分割领域关注的一个重点。针对红外图像的特性及传统分割算法效率和精度不足的缺陷,提出一种基于两种优化策略博弈的马尔可夫随机场红外图像分割方法。首先通过OTSU算法对图像进行初始分割;然后利用马尔可夫随机场理论建立图像分割模型;最后,通过SA、ICM优化策略间的博弈对图像进行分割,将两种优化策略视为博弈的两个局中人,通过寻找博弈的纳什均衡点来实现分割;实验结果表明,算法能够无人工干预地准确提取电路板红外图像所有芯片发热区域,并且很好地抑制噪声,准确处理边缘信息,具有一定的实用性和鲁棒性。Extracting the infrared image chip exothermic area at a circuit board, which is an important part of the fault detection system, has also been a key point in field of image segmentation. With the characteristics of infrared image and insufficient of traditional segmentation algorithm efficiency and accuracy considered, an infrared image segmentation algorithm based on MRF combined with two- algorithm game is proposed. First, OTSU method is used for initial segmentation, then the image segmentation model is established through MRF; Finally, segmentation is achieved through game between SA and ICM optimization strategy and by searching for the Nash equilibrium point. The simulation results show that the algorithm can not only extract accurately exothermic areas of all circuit board chips without manual intervention, but also make the speckle noise suppressed and keep the detail characteristics effectively. And the algorithm has a certain practicality and robustness.

关 键 词：图像分割 马尔可夫随机场 SA ICM 博弈论 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]