Font Size: a A A

Research For Round Wire Surface Defects Image Recognition Based On Machine Vision

Posted on:2018-12-25Degree:MasterType:Thesis
Country:ChinaCandidate:P Y YangFull Text:PDF
GTID:2348330536459990Subject:Mechanical engineering
Abstract/Summary:PDF Full Text Request
High precision round wire are main products,that surface has a variety of defects due to various reasons,which lead to low percent proficiency,in a certain company of steel wire,in Dongguan.Now,in order to accomplish work,the company detects surface defects of wire rod reliing on traditional manual and further to determine whether it is qualified,which is inefficient.With the improvement of industrial technology,the quality requirements of wire rod is higher and higher year by year,forcing the enterprise to make industrial upgrading to meet customer demand,enhance the competitiveness.According to current situation,a on-line detection system of round wire surface defects is developed,based on machine vision technology.Therefore,this paper first intends to conduct the recognition research surface defects of wire rod and lay a cetain foundation for the follow-up research.First,this paper introduces the production current situation in company and research status about machine vision technology at home and abroad;Then,in accordance with the information,including the data of round wire,parameter and surroundings collected at the production site of enterprise,characteristics and causes from six common kinds of round wire surface defects are summarized.In the meantime,peculiarities,properties,advantages and disadvantages of image acquisition hardware parts are introduced in detail.Afterwards,three common types of image noise are analyzed.And image preprocessing is completed in matlab combined with multiple filtering algorithms.According to indicators of image quality assessment,the corresponding PSNRs are analyzed and compared.Immediately following,three kinds of feature descriptors,including sift,hog,surf,are analyzed contrastively and features of image are extracted.Finally,defect images are classified and recognized by CNN combined with multilayer neural networks,preparing for the division of late wire quality grades.The results of this research can provide a reference for solving the problem of surface defects detection of wire rod.
Keywords/Search Tags:machine vision, image filtering, feature extraction, recognition and classification
PDF Full Text Request
Related items