基于多特征判别的密封圈缺陷检测  

Seal Ring Defects Detecting Method Based on Multi-features

作　　者：黎少辉[1,2] 蔡利梅[2] 

机构地区：[1]徐州工业职业技术学院,江苏徐州221140 [2]中国矿业大学,江苏徐州221116

出　　处：《煤矿机械》2015年第12期278-280,共3页Coal Mine Machinery

基　　金：江苏省"青蓝工程"资助项目(QLGC-2013-04)

摘　　要：针对小直径任意形状密封圈的缺陷自动检测,提出一种基于图像处理的多特征检测识别方法。对密封圈图像进行二值化、滤波、轮廓提取与跟踪等预处理,获取密封圈边界内外轮廓线;通过定义轮廓曲线段偏移概率矩阵,计算了偏移二阶矩、熵以及最小外接矩形宽长比,构建了特征向量;采用基于均值向量和标准差向量的分级筛选,判断缺陷是否存在及其位置,并对检测到的缺陷点按欧氏距离进行合并。实验仿真结果表明,该方法能够检测出不同形状密封圈中毛边、缺损等缺陷,多特征的组合使用使整体检测率达到91.6%,给密封圈的质量检测提供了一种良好的思路。In order to detect defects of small and arbitrary-shaped seal rings, a method based on multifeatures was proposed. The image segmentation and filtering were applied to seal ring images, and the contours were extracted. Shifting probability matrix of every contour curve segment was defined. The feature vector consisted of three components： shifting second moment and entropy based on shifting probability matrix, the wide and long ratio of minimum enclosing rectangle. The feature vectors were compared with their mean vector hierarchically to detect the defects and their positions. These points would be merged if they were adjacent. Experimental results show that the method can detect the burrs and defects of arbitrary-shaped seal rings; the recognition ratio is up to 91.6% because of the use of multi-features.

关 键 词：密封圈质检 缺陷检测 图像处理 多特征 偏移概率矩阵 

分 类 号：TQ336.4[化学工程—橡胶工业]