The Study Of Fiber Surface Defects Detection Base On Machine Vision Technology

As the backbone of modern communication network,optical cable,bearering massive traffic and stringent in reliablility requirements,its surface quality has arisen widely concern.Optical cable surface defects are kinds of tiny imperfections that occurred during the process of optical cable production,due to the deficiency of manufacturing facilities and procedures.Those defects,such as holes,bubbles,scratches,pits,discolorations,etc.,will become the sources of potential fault points,which affect system performance and damage product commercial value.Nowadays,the traditional manual quality check method can hardly fufill the precision and efficiency requirements of everincreasing scale of optical cable production.Therefore,an alternative solution is urgent needed by manufacturing enterprises.In recent years,with the improvement of imaging technology,computer technology and image processing technology,replacing eyes with machine in product vision inspection has become an important trend in modern industry.In this paper,a systematic study for the key technologies in optical cable surface defects detection based on machine vision is given and carry out the following works:(1)A method was proposed for image background removing and optical cable diameter monitoring base on boundary model and grayscale projection.Firstly,through the acquisition of radial projection image the area for optical cable can be determined,then using boundary detection algorithm refined boundary model,finally realizes the function of image background removal and cable diameter monitoring.(2)Studies for the optical image enhancement,filtering and denoising algorithm.Including image details enhancement based on homomorphic filter,image filtering based on morphological openclose reconstruction and adaptive image denoising algorithm based on median filter.(3)A method was proposed for optical cable surface defects segmentation under a textured background.Traditional image gradient operator(such as Sobel,Laplace,Prewitt,Canny operator)were high in edge detection rate,but easily affected by noise.Single threshold segmentation algorithm(such as bimodal,P parameters,Otsu)is suitable for large block segmentation of a certain image,but inefficiency for local detail detection.Multiple threshold segmentation algorithm(such as Wellner adaptive threshold)is sensitive to image local graylevel changing,but complex in computing.In this paper,a segmentation method based on watershed algorithm combined with edge detection and threshold segmentation was proposed,shows a good performance in application.(4)Studies for the recognition of optical cable surface defects.Using multi-scale LBP(local binary pattern)algorithm to extract texture feature,blending with geometry and greysacle characters as the input information of classifier.Using SVM(Support Vector Machine)to determine the pattern of cable surface defects and demonstrated to be efficient on testing data.(5)Realizing primary functions of optical cable surface defects detection system in Visual c ++ integrated development environment.Capable in common defects detection during the process of optical cable production.Obtain processing speed of 100 frames/s through parallel processing and meet the demand of real-time production monitoring.