Research Of Steel Ball Surface Face Defect Detection Technology Based On Machine Vision

With the improvement of key hardware equipment such as optical lens and industrial camera as well as the development of image processing technology,machine vision technology has been rapidly developed and applied.It has been widely used in the field of industrial products automatic detection,what is more,and has become an important research direction of steel ball surface defect detection.Steel ball surface defect detection,as an essential detection procedure of steel ball production,is played a pivotal role in steel ball surface quality.Steel ball surface defect detection technology based on machine vision has been studied,and the subject comes from steel ball company committed project.The research object is steel ball surface defects,research purpose is accurate detection of steel ball surface defects and sorting by machine vision technology,research core is edge detection algorithm for steel ball surface defect image,and research result is the design of steel ball surface defect detection system based on machine vision.Based on the research and analysis of edge detection operator,meanwhile,aiming at the deficiency of Canny operator,a self-adapting edge detection algorithm based on signal to noise ratio(SNR)with between-class variance and interclass variance is proposed.The experimental results show that the edge of steel ball surface defect image is extracted effectively via proposed image edge detection algorithm,designed steel ball surface defect detection system is stable and reliable,and is able to reach detection requirements of company.The main contents of this paper are as follows:(1)The design of steel ball surface defect detection system.Steel ball surface defect detection system is divided into hardware system and software system.Hardware system mainly includes the selection of optical lens,industrial camera,data acquisition card,lighting form and so on,and software system consists of interface design,image acquisition,region of interest selection,edge detection and extraction,sorting and display,etc.(2)A self-adapting edge detection algorithm based on SNR with between-class variance and interclass variance is proposed.Firstly,the image is filtered via the proposed adaptive median filter based on maximum filter window according to SNR.Then,the gradient amplitude is calculated in the 0°,45°,90°,135° directions,and it is suppressed with non maximum value via the 3×3 neighborhood.Finally,the high and the low threshold values are adaptively obtained applying maximum ratio of between-class variance and interclass variance,and the edge information is detected and linked with the high-low threshold values in the 3×3neighborhood.The experimental results show that the textual filtering method improvesadaptivity,denoising effect and image details preserving are better,the mean square error(MSE)is decreased,and the peak signal to noise ratio(PSNR)is increased.The textual edge detection algorithm has high precision of defect edge location,powerful self-adaption,high comprehensive evaluation of edge detection,short detection time,can effectively extract the edge of steel ball surface defect image.(3)Steel ball surface defect detection system verifies reliability and accuracy.Through orthogonal test analysis,the dependability and precision of the system is verified,the primary and secondary influence and superior combination that circumferential unfolding motor speed,reciprocating unfolding motor speed,and the number of each steel ball image acquisition respectively for error rate and detection efficiency have been studied.