Research And Realization Of Metal Surface Defect Detection System

Product quality inspection is crucial for the healthy development of an industry.For metalwork,surface inspection is a matter of cardinal significance.It’s a necessity for both metalwork quality control and performance assessment.In our country,surface disfigurements of metal products are detected mainly by eyes,which results in low efficiency,inaccuracy,and inconsistency.Such method can hardly answer to the demand of a modernized and fast-growing industry or compare with the real-time quality inspection system of developed countries.A more prosperous industry with quality products calls for an automated and intellectualized metal surface inspection system.In this regard,machine vision comes to its need.This study mainly deals with the principles and implications of metal surface inspection technology.The research aims are as follows:(1)To meet the specific requirements of metal surface inspection,a hardware design based on deep learning technology is proposed borrowing the existing related technologies from home and abroad.(2)A more efficient and accurate algorithm for automated inspection system is proposed as opposed to the traditional one.(3)The hardware and software mentioned above are combined into the new automated metal surface inspection system with deep learning technologyThe structure of this study is as follows:(1)It systematically combs through the existing technologies in metal surface inspection from home and abroad.A device for the highly functioning visual system is selected based on the specific needs of metal surface inspection.Besides,the device is also designed to build an online automated metal surface inspection system based on machine vision to improve the efficiency and accuracy of the inspection as required.The process involves creating the stereo image of the surface by cooperative sampling with three cameras and LED automatic strobe light source,which provides sufficient source for the detection of surface disfigurements.(2)It focuses on the technological principles of existing metal surface inspection systems by discussing their strengths and weaknesses.Thus,an improved algorithm on the basis of morphology is proposed to improve the inspection technology.Besides,the periodic texture is removed by CNN technology instead of optical filter because of the possible interference from uneven illumination and spatial structure.Through the scientific and comprehensive experiments,the anti-noise and separation performances of the method employed in this study are assessed.(3)The software architecture is designed for the functioning system based on Visual Studio,Accord.net and so forth.Thus,with the integration of both hardware and software,the metal surface inspection system with machine vision at its core is established,which is an innovation in metal surface inspection.