Roller Bearing Surface Defect Recognition Method And Software Design

As the foundation of machinery,bearing is an important part of developing modern industry.The geometrical precision and surface defects of the roller directly affect the performance of the bearing.At present,most bearing roller manufacturers use traditional manual visual inspection or some non-destructive testing methods based on physical characteristics to detect the surface defects of the roller.Machine vision detection has become the mainstream of surface defect detection,but there are still the following shortcomings:low accuracy of image feature segmentation,low efficiency of image processing methods,less kinds of defects in detection and other problems.Based on this,this thesis designs an accurate identification method for matching the defect features obtained by cooperative segmentation of gray scale value,gradient difference threshold and background discrimination with the defect images in the database,and designs a defect image database and a surface defect detection software system.The main research contents are as follows:First,combining threshold segmentation method and region extraction method,a new method of surface defect segmentation and identification of bearing roller is designed.Firstly,spatial information and range information are taken into account in the filtering method,and the measured image is preprocessed by bilateral filtering with accelerated optimization to smooth the noise and protect the edge information of the defect target.Secondly,the image is divided into grids,and the image blocks are segmented into gray scale value,gradient difference threshold value and adaptive background discrimination statistical analysis.The connected region marking method of the image is used to integrate the image blocks containing defects into complete defect features to complete the segmentation of image defect features.Finally,the morphological classification of the segmented defect features was carried out and the correctly classified defects in the defect database were matched based on the similarity of image Hu invariant moment to accurately identify the defect information.Secondly,this thesis establishes a whole set of defect detection software system and defect image database.The software part adopts modular design as required.The design process of defect detection software,the design of main modules and the database development part are analyzed and verified in detail and the test is completed.In addition,the defect database containing defect feature information is established according to the algorithm requirements.Thirdly,the algorithm in this thesis is analyzed in a confirmatory experiment.Based on the images to be tested collected by the experimental platform,the step experiment of defect recognition and the comparison experiment of common segmentation algorithms are completed.The effect of defect feature recognition is qualitatively analyzed.The effect of surface defect identification of bearing roller is better.The defect identification method designed in this thesis is universal and has important reference value for the research of workpiece surface detection technology.