基于机器视觉的金属零件表面缺陷检测研究  

A Research on Surface Defect Detection of Metal Parts Based on Machine Vision

作　　者：孙姿姣 罗芳[1] 李阳辉[2] SUN Zijiao;LUO Fang;LI Yanghui(School of Mechatronics and Automotive Engineering,Qingyuan Polytechnic,Qingyuan 511510,China;School of Infommation Technology and Creative Arts,Qingyuan Polytechnic,Qingyuan 511510,China)

机构地区：[1]清远职业技术学院机电与汽车工程学院,广东清远511510 [2]清远职业技术学院信息技术与创意设计学院,广东清远511510

出　　处：《清远职业技术学院学报》2025年第1期42-48,共7页Journal of Qingyuan Polytechnic

摘　　要：目前制造业中,金属零件的缺陷问题会导致重大经济损失,主要问题在于零件缺陷小且缺陷位置出现随机,传统人工检测难以区分微小缺陷位置与非缺陷位置,且人力成本高,经济效益低下。针对这一问题,研究提出一种基于机器视觉的金属零件表面缺陷检测方法,通过机器视觉检测代替人力劳动,同时采用交互式空间位置注意力模块,解决了金属零件表面的缺陷不明显难以检测的问题,采用对偶局部-全局Transformer模块,解决了缺陷区域与周围正常区域难以区分的问题,提高了金属零件表面微小缺陷的检测性能,从而提高企业经济效益。In the current manufacturing industry,the occurrence of defects in metal parts can result in substantial economic losses.The primary challenge lies in the small size and random distribution of these defects.Traditional manual detection struggles to differentiate between defective and non-defective areas due to their subtle nature,leading to high labor costs and low economic benefits.To solve this issues,a method for surface defect detection of metal parts based on machine vision is proposed.Machine vision detection replaces human labor,while an Interactive Spatial Position Attention Module is employed to tackle the difficulty of detecting inconspicuous surface defects on metal parts.Additionally,a Dual Local-global Transformer Module is utilized to enhance the distinction between defect areas and surrounding normal areas,thereby improving the detection performance for small surface defects on metal parts and ultimately enhancing enterprises’economic benefits.

关 键 词：机器视觉 缺陷检测 交互式空间位置注意力模块 对偶局部-全局Transformer模块 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]