纹理表面缺陷机器视觉检测方法综述

Review of Machine Vision Detection Methods for Texture Surface Defects

纹理表面缺陷检测在机器视觉领域具有意义和挑战性,其历史可以追溯到20世纪中后期,近年来随着深度学习技术的蓬勃发展,纹理表面缺陷检测技术大幅飞跃。直至今日,关于纹理表面缺陷检测的调研和综述仍然很少。在此背景下,本文回顾2017年-2021年间200余篇纹理表面缺陷机器视觉检测论文,对纹理表面缺陷机器视觉检测研究进展进行了及时、全面的调查;分析了纹理表面缺陷检测的发展历史和最新研究进展,原则上将纹理表面缺陷机器视觉检测方法分为传统方法与深度学习方法,并对二者进行了深层次研究分析,特别是深度学习方法;对近期出现的几种纹理表面缺陷机器视觉检测方法主题进行总结的同时,也对这些主题的研究进展进行了综述。最后,对未来的研究趋势进行了展望,以期为后续研究提供指导和启示。

Texture surface defect detection is meaningful and challenging in the field of machine vision.The history of texture surface defect detection can be traced back to the middle to late 20th century.Moreover,in recent years,with the flourishing development of deep learning technology,texture surface defect detection technology had a big leap.However,so far,there are still few surveys and reviews of texture surface defect detection.Against such a background,we comprehensively reviewed more than 200 papers about texture surface defect detection with machine vision from 2017 to 2021 and made a timely and comprehensive investigation of its research progress.This paper reviews the development history and latest research progress of texture surface defect detection.In principle,the methods of texture surface defect detection by machine vision are divided into the traditional method and the deep learning method,which were studied and analyzed deeply,especially the deep learning method.The paper summarizes several methods of texture surface defect detection by machine vision that appeared recently and reviews the research progress of these methods.Finally,it introduced the future research trends to provide enlightenment for further studies.