基于图像特征识别的部件缺陷无损检测方法(英文)  

Image Feature Identification Based Method in Nondestructive Test of Defective Parts

作　　者：谭显坤[1] 杨小义[2] 

机构地区：[1]重庆交通大学应用技术学院,重庆400074 [2]重庆师范大学教育科学学院,重庆400030

出　　处：《机床与液压》2012年第12期69-74,共6页Machine Tool & Hydraulics

摘　　要：图像处理在金属部件内缺陷无损识别检测中有重要应用,但在原始图像辨识处理中至今仍存在如边缘不清晰等诸多问题。为消除传统图像辨识中的边缘不连续,伪边缘、过分割,提出了一种基于模糊形态学的图像几何特征辨识算法。该算法基于模糊形态学辨识结构元素特征,分析各元素间的相互关系,对经预处理后的图像确定隶属度函数参数,采用最优阈值自动识别方法确定图像最佳分割阈值,以最佳阈值和路径代价函数为约束条件,缩小路径搜索范围,提高算法执行速度,借助分水岭算法重现图像边缘,辨识出图像的几何特征。最后以一种特殊的图像几何特征识别为例进行了数字仿真。仿真图像显示,经处理后的图像边缘连续,无伪边缘和过分割现象,图像更加平滑,柔和性更好。研究结果表明,该方法是可行和有效的。Image processing has very important application in nondestructive test of defective metal parts, but there exist many problems in the process for original image identification processing, such as edge unclear. In order to remove edge discontinuousness, forge edge and over-seg- mentation, the paper proposed an algorithm of fuzzy morphology based feature identification in image processing. The method based on structure element of fuzzy mathematical morphology to identify feature, analyzed elements correlation, decided subjection function after preprocessed images, ascertained best segmentation threshold adopting technology of auto-recognition best threshold. Then the algorithm took best threshold and path cost function as constraint condition to reduce search scope, enhance algorithm executing speed. Finally it extracted edge and identified image geometry features by dint of watershed algorithm. The paper took a special kind of recog- nition of image geometry features as an example to make the digital simulation, and the pro- cessed images demonstrated that it not only could make image edge continuous, but also could remove the phenomena of forge edge and over-segmentation, and the image change smoother and more flexible by using the algorithm to process images. The research result shows that the proposed method is feasible and effective.

关 键 词：模糊形态学 图像辨识 分水岭算法 几何特征提取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]