基于模糊熵和模拟退火算法的双阈值图像分割  

Dual-threshold image segmentation based on fuzzy entropy and simulated annealing algorithm

作　　者：郑毅[1,2,3] 郑苹[4] 

机构地区：[1]山东工商学院信息与电子工程学院,烟台264005 [2]西安电子科技大学技术物理学院,西安710071 [3]山东工商学院智能信息处理山东省高校重点实验室,烟台264005 [4]华中科技大学图像识别与人工智能研究所,武汉430074

出　　处：《电子测量与仪器学报》2014年第4期360-367,共8页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(60970105;61173173;61272430);山东省自然科学基金(ZR2012FL09;ZR2013FM015);山东省住房和城乡建设厅科技项目(2011YK060);山东省高等学校科研计划(J11LG12)

摘　　要：在火炮身管弯曲度测量系统中,为了能够从光靶图像中同时提取标定图案和激光光斑,提出了一种双阈值图像分割方法。结合模糊数学理论和最大模糊熵判据,把光靶图像中的像素灰度级分为黑、灰和亮3个模糊子集,用于畸变校正的标定图案的像素灰度级隶属于黑模糊子集,用于测量的激光光斑的像素灰度级隶属于亮模糊子集。使用模拟退火算法确定最优的模糊熵参数组合,降低了计算复杂度,并且最大模糊熵判据只含有4个模糊熵参数,减小了搜索空间。对光靶图像进行了双阈值分割实验,并与最大类间方差双阈值法进行了比较。实验结果表明,所提方法能够自动、有效地选取双阈值,分割效果优于最大类间方差双阈值法。A dual-threshold method is proposed to extract a calibration pattern and a laser spot from a target image simultaneously in a gun barrel camber measurement system.Using fuzzy mathematics theory and a maximum fuzzy entropy criterion,the proposed method can classify the target image into three fuzzy subsets:dark,gray and bright fuzzy subset by their gray levels.Gray levels of the calibration pattern used for distortion correction belong to the dark fuzzy subset,and ones of the laser spot for measurement belong to the bright fuzzy subset.A simulated annealing algorithm is implemented to search an optimal combination of fuzzy entropy parameters,which has a low computational complexity.And there are only four fuzzy entropy parameters in the proposed method,and search space is in small size.The proposed method is tested and compared with an Otsu's dual-threshold method.The experimental results show that the proposed method can determine dual-threshold automatically and efficiently,and has a better segmentation than the Otsu's dual-threshold method.

关 键 词：图像分割 阈值 隶属度函数 模糊熵 模拟退火 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]