基于改进水平集方法的腐蚀材料图像特征提取  

Feature extraction from corrosion material images via improved level set method

作　　者：彭丽丽[1] 吴萍萍[2] 

机构地区：[1]苏州科技大学外国语学院,江苏苏州215009 [2]重庆理工大学计算机科学与工程学院,重庆400054

出　　处：《苏州科技学院学报（自然科学版）》2016年第2期45-50,共6页Journal of Suzhou University of Science and Technology （Natural Science Edition）

基　　金：原国防科工委基金资助项目(H102006A00)

摘　　要：由于良好的稳定性和拓扑无关性,水平集已被广泛应用到图像分割中。针对材料腐蚀的锈点特征,采用无需重新初始化的水平集方法,结合不同尺度空间理论的特征检测方法,对边界追踪函数引入Dog算子与Log算子结合以强化边缘检测极值点,提取材料图像的腐蚀锈点。通过理论分析和仿真实验验证表明,当迭代次数过低时,未改进算法对材料腐蚀锈点图像特征提取容易出现过度分割,改进后的算法性能相对稳定,能够获得较好的提取效果。此外,在获得同等分割效果时,改进后算法的速度相对提高了22.9%。Level set has been widely applied to image segmentation because of its good stability and topological independence. For the rust spot features of material corrosion ,we adopted the level set method without initializing and introduced the Log operator and the Dog operator to the boundary tracking function to strengthen the edge detection extreme value point with the feature detection method of different-scale spacial theory. Based on all this,we extracted the corrosion rust spots of the material images. Through theoretical analysis and simulation ex-periments,the results show that the unimproved method is prone to excessive segmentation in the feature extrac-tion of corrosion rust spots when iterations are too low ,while the improved method is more stable and takes better extraction effect. In addition,the improved method has sped up by 22.9% when the segmentation effects are the same.

关 键 词：水平集 腐蚀锈点 图像分割 特征提取 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]