基于机器视觉的轴承跑合带检测方法研究

Research on the evaluation of running-in belt quality of bearings based on machine vision

针对轴承跑合带图像采集流程繁琐、人工检测任务重等问题,提出搭建一套基于机器视觉的轴承跑合带检测系统,以实现跑合带的批量自动化检测需求。该系统主要包括零件运输、图像采集模块,并配有图像处理算法。为适应不同类型轴承跑合带的检测需要,该系统将检测分为跑合带分割与评估和跑合带缺陷检测两方面。对于跑合带分割与评估,系统选用FCN8x网络作为识别算法,自动完成对轴承跑合带图像的识别与分割,并结合其深浅变化的均匀性及宽度变化的连续性因素对其进行评估;对于内部缺陷检测,系统使用基于边界跟踪的轮廓提取对缺陷进行判别与分类。通过对300组不同样式的轴承跑合带样本进行测试,验证该系统对多样式型号轴承均具有良好的检测效果。

This paper presents a machine vision based system for the automated evaluation of running-in belt quality of bearings,addressing issues related to image acquisition and labor-intensive manual detection.The system comprises three essential modules:part transportation,image acquisition,and image processing.To apply to various types of bearing running-in belts,the system employs a two-step approach:running-in belt segmentation and defect detection.For running-in belt segmentation,FCN8x was taken as the recognition algorithm.It automatically identifies and segments images,evaluating the results based on color change uniformity and width change continuity.For defect detection,contour extraction using boundary tracking is employed to accurately identify and categorize defects.Empirical testing on 300 sets of bearing running-in belt samples of varying styles attests to the system's robust detection capabilities across a diverse range of bearing models.This system streamlines the detection process,enhancing efficiency and accuracy in bearing running-in belt quality evaluation.