基于熵率聚类分割和环带差分的油封缺陷检测

Entropy Rate Clustering and Belt Difference for Oil Seal Defect Detection

针对油封工件表面啃伤缺陷与背景对比度较低,检测区域灰度分布不均,区域分割与缺陷提取困难等问题,提出了基于熵率聚类分割和环带差分的油封缺陷检测方法。首先将油封图像进行超像素分割,分割为多个检测区域。根据油封灰度值轴向差异大、环向分布均匀的特点,采用环带均值背景差分方法检测啃伤缺陷。实验选择黑色橡胶骨架油封测试,结果证明,该检测算法能够有效实现油封区域分割和表面啃伤缺陷检测,区域分割正确率约98%,缺陷检出率可达95%。

Lowcontrast of oil seal surface grawdefect and background,uneven distribution of gray level in detection region,makes the region segmentation and defect extraction extremely difficult. Therefore,an oil seal defect detection method based on entropy rate clustering segmentation and ring belt difference is proposed,the seal surface is divided into a plurality of detection area through super pixel segmentation.Then,the properties of inner/outer belt regions are analyzed. The pixel value of the regions are homogeneous in circumferential direction and variation in axial direction,the belt difference is proposed for the defect detection. Experiments were conducted on black rubber skeleton oil seal. The results showthat the proposed detection algorithm can effectively segment the area of oil seal and detect surface grawdefect.The accuracy of area segmentation is about 98%,and the detection rate of defects can reach 95%.