基于改进Faster R-CNN的轴承端面缺陷检测方法  

Bearing End Face Defect Detection Method Based on Improved Faster R-CNN

作　　者：李晓宇 张涛[1,2] 刘宏伟 汪勇 LI Xiao-yu;ZHANG Tao;LIU Hong-wei;WANG Yong(Beijing Key Laboratory of High Dynamic Navigation Technology,Beijing Information Science&Technology University,Beijing 100101,China;Key Laboratory of Modern Measurement&Control Technology,Ministry of Education,Beijing Information Science&Technology University,Beijing 100101,China;Institute of Remanufacturing Industry Technology,Jing-jin-ji,Cangzhou Hebei 062450,China)

机构地区：[1]北京信息科技大学高动态导航技术北京市重点实验室,北京100101 [2]北京信息科技大学现代测控技术教育部重点实验室,北京100101 [3]京津冀再制造产业技术研究院,河北沧州062450

出　　处：《计算机仿真》2025年第1期491-496,共6页Computer Simulation

基　　金：国家自然科学基金重大科研仪器项目(52227804);国家自然科学基金面上项目(52274003);北京市教育委员会科学研究计划项目资助(KM202111232004)。

摘　　要：针对轴承端面小目标缺陷识别率不高、漏检多等问题,提出一种基于改进Faster R-CNN的轴承端面缺陷检测方法。首先,自主采集了轴承端面缺陷数据集,并利用Mixup、Mosaic和Getrandom三种方式对数据集进行扩增,保证数据样本充足,有效避免了过拟合问题;其次,对比了ResNet50和VGG16两种不同特征提取网络,选择检测精度更优的ResNet50作为主干网络,并结合特征金字塔(FPN)网络,以提升网络训练性能;最后,针对原有算法锚框过大的问题,使用K-means算法改进锚框方案,提高小目标缺陷检测能力。实验结果表明,改进Faster R-CNN算法的平均精度均值(mAP)可达85.80%,具有更好的检测精度和较低的漏检率,以上研究为轴承端面缺陷检测系统的改进提供了新思路和方法,对工程实践具有重要意义。Aiming at the problems of low recognition rate and high omission rate for small target defects on bearing end faces,a bearing end face defect detection method based on an improved Faster R-CNN is proposed.Firstly,a bearing end face defect dataset is collected autonomously,and the dataset is augmented using three methods,namely Mix up,Mosaic,and Get random,to ensure sufficient data samples and effectively avoid overftting.Secondly,two different feature extraction networks,ResNet50 and VCG16,are compared,and ResNet50,which has better detection accuracy,is selected as the backbone network.Feature Pyramid Network(FPN)is also combined to improve network training performance.Finally,to address the problem of anchor box size being too large in the original algorithm,the K-means algorithm is used to improve the anchor box scheme and enhance the detection ability of small target defects.Experimental results show that the mean average precision(mAP)of the improved Faster R-CNN algorithm can reach 85.80%,with better detection accuracy and a lower omission rate.This research provides new ideas and methods for the improvement of bearing end face defect detection systems,and is of great significance to engineering practice.

关 键 词：轴承端面 缺陷检测 数据增强 多尺度融合 聚类算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]