Research On Key Technology Of Surface Defect Detection In Small Forgings(Pipe Joint)

The defects of surface of small forgings are caused by some factors,which are forging process and related forging equipment during processing,then the appearance and benefits of products are affected.Human inspection with eyes detection of surface defects of small forgings have the disadvantages of low efficiency,high cost,high labor intensity and missed inspection and cannot satisfy the development demand of forging industry.Hence,it is crucial how to effectively and accurately the defects of surface of small forgings in the production process for improving the quality of surface of small forgings.With the rapid development of information science technology of late years,machine vision provides the foundation for detecting the defects of surface of small forgings.In this thesis,does take pipe joints as the studying object,then the image preprocessing algorithm,image initial detection algorithm,image segmentation algorithm,characteristics extraction and selection algorithm,and pattern recognition algorithm of small forging surface defect images are studied.The overall plan of the detection of surface defects is designed.Mainly accomplished the research contents as followings:1.The characteristics of surface,process of production,comments on the defect types,which causes and features has been analyzed and the technical indicators of the detecting system to the pipe joints are introduced,and the inspection system is overall designed,both software and hardware.2.The foreground extraction algorithm,preprocessing algorithm,image initial detection algorithm,image segmentation and defect region target location algorithm on the surface defect image of the pipe joints are studied.The GMM algorithm is used for extracting the foreground region of the image;the surface defect image of the pipe joints is preprocessed by the adaptive median filtering and the piecewise linear transformation;the local binary pattern detection algorithm determine the image acquired of pipe joints whether the existence of defects;the OTSU algorithm is used to the defect image segmentation on pipe joints surface;the area where the defect islocated according to the coordinates of the horizontal projection and the vertical projection of the defect area,and the segmentation and positioning of the defect area of the pipe joints are realized.3.The feature extraction and selection algorithm of pipe joints defects area are studied and geometric features,texture features,gray scale features and invariant moment features of the pipe joints defect area are extracted.The principal component analysis method is used for selection features to realize debasing dimension of features data.4.The classification and recognition algorithm of the surface defect image of the pipe joint is studied,meanwhile,the multi-class classifier based on the BP neural network for the surface defect image of the pipe joint is designed.Tested by 90 test samples,the accuracy of classification recognition was above 95%.