基于机器视觉的轴承表面微小缺陷检测技术

Slight Defects Detection Technology of Bearing Surface Based on Machine Vision

轴承表面缺陷检测要求准确率较高,因此研究基于机器视觉的轴承表面微小缺陷检测技术。使用CMOS工业相机采集轴承表面图像,利用线性拉伸法与中值滤波方法预处理轴承表面图形,提升图像的亮度并消除干扰信息,使用八连通区域标记法标记处理后的轴承表面图像,获得连通区域数量,使用该数量计算得出缺陷特征的面积、长宽比、周长等几何特征,将这些特征作为剪枝分类决策树的输入,经过训练后,输出轴承表面微小缺陷分类结果,实现轴承表面微小缺陷检测。试验结果显示,该方法预处理后,图像质量较高,能够检测出轴承表面多种微小缺陷,并且检测结果准确率较高。

Bearing surface defect detection requires high accuracy,so the bearing surface small defect detection technology based on machine vision is studied.Using CMOS industrial camera image acquisition bearing surface,using the method of linear stretch preprocessing bearing surface graphics with median filtering methods,improve the brightness of the image information and eliminate interference,using eight connected area method of marking the bearing surface of the image after processing,get connected area number,use the number calculated the area of the defect feature,aspect ratio,perimeter,such as geometric features.These features are used as the input of pruning classification decision tree.After training,the classification results of bearing surface minor defects are output to realize bearing surface minor defects detection.The test results show that the image quality of the method is higher after preprocessing,and it can detect a variety of small defects on the bearing surface,and the detection accuracy is higher.