采用机器视觉检测方法对深沟球轴承装配过程中的滚珠遗漏缺陷进行自动检测。引入3种光源照明方案用于采集轴承图像,采用中值滤波去除图像噪声,基于圆形Hough变换和极坐标展开方法进行轴承图像的圆形检测和矩形展开。选用完好轴承80个、滚珠遗漏轴承60个进行实验。结果表明:采用背光配合同轴光的照明方式可有效减少轴承表面反光;采用中值滤波对图像进行预处理,可以在消除图像孤立噪声点的同时,使图像少一些模糊;采用圆形Hough变换可以快速获取轴承的内外环图像并对其进行定位,然后通过笛卡尔极坐标展开方法将经过预处理的深沟球轴承图像归一化展开成矩形,最后通过设置灰度阈值实现滚珠缺漏位置的检测和识别。本文方法对80个完好轴承的识别率为92.5%,对60个滚珠遗漏轴承的识别率为93.3%。

The roller missing defect in the assembling process of deep groove ball bearings is detected automatically with the machine vision method. Three lighting schemes are presented for acquisition of bearing images. The image noise is removed by median filter. The Hough transform algorithm and the polar coordinate expansion method are used for circular detection and rectangular expansion of bearing images. 80 perfect bearings and 60 roller missing bearings are selected for test. The results show that the lighting system of coaxial light source combined with backlight can effectively reduce the surface reflection of bearings. The pre-processing of the image with the median filter can eliminate the isolated noise and make the image less vague. The Hough transform algorithm can quickly obtain the images of bearing inner and outer rings and locate them. The Cartesian polar coordinate expansion method can expand the roller bearing images into rectangles. The detection and recognition of the roller missing position are realized by setting the gray threshold. The recognition rates of the proposed method for 80 perfect bearings and 60 roller missing bearings are 92.5% and 93.3%, respectively.