Research On Surface Defect Detection System Of End Plug Based On Machine Vision

As an important part of energy fuel,the End plug's apparent quality control is the key to improve the performance of the product and improve the production technology.With the continuous development of computer science,digital image processing,intelligent control and other technologies,based on machine vision defect detection has gradually replaced the traditional manual inspection method,its low cost,no damage,fast and accurate characteristics,so that the technology widely used in industry areas.According to the characteristics of the end plug for the cylindrical workpiece and the metal material,the end plug is detected by Multiple platforms and regions.Firstly,the appearance characteristics of the end plug are analyzed.According to a variety of defects on the end plug's surface,we describe the working principle of the detection system and put forward the end plug defects for multiple platforms detection.At the same time,we do a series of hardware design for the end plug detection system.According to the requirement of the software,we put forward the scheme in two aspects: the design of interactive interface and the design of software flow.Secondly,due to the high reflectivity of the end plug surface,there is a problem that the optical illumination is uneven,and the defect detection rate is not high.Therefore,the image is preprocessed.In this paper,the heterogeneous diffusion model is applied to the surface image of the end plug to denoise,which is facilitate the subsequent binarization processing,the transition region extraction and the region of interest extraction.In this paper,based on the Hough transform theory,an improved random Hough transform method is proposed to obtain the set of pixels at the edge of the inner hole.Using the circular fitting method and the residual analysis,the invalid points in the set of cyclic elimination points finally obtain the center point and radius,and detect the defects at the same time.To effectively detect the end plug defection such as scratches,bumps,indentations,stains and other defects,the core algorithm of LBP operator and its uniform pattern LBP operator,the rotation invariant LBP operator,multi-scale LBP operator and LBP operator and the variance of the various methods are analyzed.a defect detection algorithm based on Multi-scale Local Binary Pattern(MSLBP)and Local Image Variance(LVAR)is proposed.The Multi-scale LBP histogram is used to reflect local graphics texture pattern and the LVAR to outstand the contrast of image intensity.Then,the weight values calculated from LVAR are used to adjust the extraction and measurement of LBP local texture.Thus,the automatic surface defect detection for end plug is achieved.Finally,the algorithm has a lot of experiments show that the surface defects of the plug are well detected,and the scope of the defect,threshold,area,size and other parameters are determined.At the same time as the transition zone,the light detecting defects of different scale texture under high robustness,the total detection rate reached 89.4%,the missing rate is0%,which satisfies the requirements of detection.It can be used to realize the automatic detection of the quality of the end plug,which has a certain guiding significance and practical value for the real-time detection of metal parts.