Research And Application Of Surface Defect Identification Method For Galvanized Steel Sheet Based On Image Processing

Galvanized sheet is a cold pressed sheet as the raw material,coated on the surface of galvanized layer,to increase corrosion resistance,prolong the service life of the steel plate.Surface quality defect is an important factor affecting the quality of galvanized sheet.The detection of surface defect of galvanized steel is always a concern of steel enterprises.The surface defects of the traditional manual inspection of galvanized sheet cannot achieve satisfactory results,in recent years,microelectronics,computer technology,automation technology,machine learning,artificial intelligence theory and practical,there is a new way to quality detection of surface defects of galvanized sheet.In this thesis,surface quality inspection and surface defect recognition and classification technology of galvanized steel sheet are studied.(1)It briefly analyzed the types,characteristics and causes of the surface defects of galvanized sheet.Three common types of typical defects were selected as the object of this dissertation.They were scratch after plating,hole defects and dark spot defects.(2)In-depth research on defect feature extraction and selection algorithm and defect classification principle algorithm is carried out,including ReliefF combined with correlation criteria,which evaluate multi-dimensional mixed weighted feature algorithm,BP neural network algorithm and XGBoost algorithm.The structure principle and advantages and disadvantages of the three algorithms are introduced,and some improvements are made in view of the shortcomings.ReliefF is used to evaluate the selection multidimensional mixed weighted feature algorithm to remove redundant features which do not work on the classification,and remove redundant features with correlation criterion.After two times of screening,the optimized feature combination is obtained.Then the classifier model is designed respectively by BP neural network and XGBoost algorithm,and three kinds of defect types are identified and classified.(3)Put three kinds of typical defects of galvanized steel surface as the experiment object,first of defect image to grayscale conversion,filtering,enhancement,defect segmentation and edge detection preprocessing using image processing technology to get a complete two value image feature extraction.Then,the geometric features,texture features and grayscale features are extracted,and the feature selection and de redundancy process is carried out by combining the ReliefF algorithm with correlation criterion.Finally,two classifiers were used to do galvanized steel surface defect classification experiment respectively,and achieved good experimental results.The successful implementation of recognition and classification of galvanized steel surface defects has high application value.