摘要
针对磁环表面缺陷图像具有对比度低、纹理背景复杂、缺陷种类多和亮度不均匀等问题,提出了一种基于改进自适应Canny算法和掩模技术的磁环表面缺陷提取方法。首先,在分析磁环表面图像中不同区域灰度特征和梯度特征的基础上,通过拟合磁环内外轮廓构造掩模图像,以便屏蔽磁环背景区域的干扰;然后,利用提出的基于8邻域各向异性滤波的改进自适应Canny边缘检测算法,抑制磁环表面纹理的干扰;最后,利用图像数字形态学增强边缘连通域,并利用构造的掩模图像提取磁环表面缺陷。利用开发的样机进行了大量的在线实验。实验结果表明,本文缺陷提取算法稳定性好,鲁棒性强,能够准确、快速地提取出磁环表面图像各区域的缺陷,表面缺陷检测的准确率为97.3%。
Aiming at low contrast,complex background texture,many kinds of defects and unev en brightness in the magnetic ring surface defect image,a surface defect extract ion method of magnetic ring is proposed based on an improved genetic algorithm of adaptive Canny algorithm of edge-detection met hod and masking technology in this paper.Firstly,based on the analysis of gray and gradient features magnetic surface image in different regions of the interfere nce of magneticing background area is shieldedby fitting the contour mask image in and out of magnetic ring.Secondly,an inproved adaptive Canny edge detection algorithm based on 8neighborhood anisotropic filtering is used for suppressing the interference of magnetic ring surface texture .Finally,the edge connected domain is enhanced by morphology using digital image,and the mask image is used to extract the magnetic surface defect.A large number of online experiments are carried out using the developed prototype.Experimental results show that the defect extraction algorithm have good stability,strong ro bustness, and can accurately and quickly extract the defect of various regions in magnetic ring surface image.And the accuracy rate of surface defect detection is 97.3%.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2017年第9期978-984,共7页
Journal of Optoelectronics·Laser
基金
国家自然科学基金(61374022)
浙江省公益性技术应用研究计划(2014C33109)
浙江省新型网络标准及其应用技术重点实验室开放课题(2013E10012)资助项目
