荧光磁粉探伤自动缺陷识别方法研究  

Automatic Flaw Recognition Method of Fluorescent Magnetic Detection

作　　者：陆宝春[1] 李建文[1] 陈吉朋[1] 王婧[1] 李纯健 汤海昌 

机构地区：[1]南京理工大学机械工程学院,江苏南京210094 [2]盛捷达探伤设备制造有限公司,江苏盐城224300

出　　处：《南京理工大学学报》2010年第6期803-808,共6页Journal of Nanjing University of Science and Technology

基　　金：国家科技型中小企业技术创新基金(09C26213203712);国家大学生创新实验计划(101022802)

摘　　要：针对荧光磁粉探伤的人工观察识别过程中工作量大、效率低、准确性低及健康危害严重等问题,提出一种新的自动化图像处理与缺陷识别方法。通过分析磁痕图像,设计了一种基于Photoshop中"色彩范围"选择功能的彩色图像分割算法;基于图像形态学提出图像去噪与还原算法;在区域生长理论的基础上进行相关优化,实现单个连通区域的高效提取;基于真伪信号的特征差异设计了缺陷识别方法。以样本图片为例,在集成了OpenCV的VC环境下进行了算法调试,结果显示:彩色图像分割模块滤除了原始图像中由于紫外灯光照射形成的蓝紫色区域和附着磁粉形成的暗绿色区域等干扰信号;开运算后图像中的斑点噪声已经完全被去除,部分细小区域在开运算过程中被断开,导致图像失真;通过闭运算图像得到了还原,图像特征指标保持良好;在识别效率显著改善的同时,识别方法的通用性与准确性也都得到了明显提高。To overcome the problems of heavy workload,low efficiency,low accuracy and serious health hazard of the process of manual observation in fluorescent magnetic detection,a new automatic image processing and flaw recognition method is proposed.Through analyzing magnetic mark photos,a color image segmentation method is designed based on the choice function of ＂color range＂ in Photoshop.The image denoising and restoration algorithm is given based on an image morphology.The single connected region is extracted efficiently through the optimization of region growing algorithm.The flaw recognition method is designed based on the character differences between true and false signals.Taking the sample image for example,the algorithm is debugged in the VC environment integrated with OpenCV.The results show： the interference signal of the blue-purple region formed by ultraviolet light irradiation and the dark green region formed by attachment magnetic powder in the origin image is flitted by the color image segmentation model;the speckle noise is removed after open operation,image distortion is caused by disconnecting some smell regions in the process of open operation;the image is reduced by close operation,the character targets of the image are kept well;with the improvement of the recognition efficiency,the universality and accuracy of the recognition method is improved.

关 键 词：荧光磁粉探伤 缺陷识别 图像分割 图像形态学 区域生长 

分 类 号：TG115.28[金属学及工艺—物理冶金]