Research On Visual Online Detection Of Surface Defects In Air-blown Microcables

The surface quality of air-blown microcables is an important part of product quality,whose detection is an important part of the product quality inspection process.Nowadays,optical cable manufacturers mainly rely on manual inspection for micro-cable surface defect detection,which has the disadvantages of low detection efficiency and high missed detection rate.In recent years,it is possible to use machine vision inspection technology to detect micro-cable surface defects automatically because of the development of it.However,it is found in the research that there exists some problems such as high detection false alarm rates,research and development based on specific production lines and so on,which affect the detection effect and promotion in the cable industry.This article focuses on the existing problems and study on the lighting system,detection algorithm,feature extraction and recognition modules in the detection system deeply,which solves the micro-cable surface with complex textures,water droplets caused by high false detection false detection rate and adaptive lighting adjustment and other issues.The main research contents of the paper are as follows:1.Optimization research on lighting system.The red and white LED light sources are selected for comparison,and the evaluation of the micro-cable image and the grayscale variance statistics determine that the red light source has advantages in collecting high-quality images.Aiming at the problem that the lighting system needs to manually adjust parameters when the test microcable is changed during the test,the communications between the computer and the lighting drive circuit are established by the single-chip microcomputer,and the detection system is used to adjust the illumination intensity through the adaptive adjustment algorithm to realize the adjustment of lighting environment adaptively,which is suitable for testing various types of micro-cables.2.Aiming at the problem of high false positive rate of defect detection due to the complex texture on the surface of the micro-cable,the edge detection algorithm was used to smooth the complex cable surface pattern to reduce the impact on defect segmentation.An edge detection algorithm scheme centered on a bi-orthogonal wavelet algorithm with symmetrical support of vanishing moment as the core is proposed,which has good performance on the micro-cable samples containing characters,water drops,damage,scratches and other tests on the surface compared with traditional edge detection algorithm,this algorithm has advantages in segmenting the target region in the complex texture background.3.Aiming at the problem of water droplets sticking to the surface of the micro-cable,which caused a high false alarm rate of defect detection,this paper selects water droplets as an example to conduct feature value extraction and identification.Eigenvalue extraction was performed on the waterdrop samples by using local binary mode features,and the principal component analysis method was used to reduce the data.The eigenvalues of every images were reduced to 20,and the feature retention rate was 95%.By building database training on waterdrop sample images and using support vector machines to train and recognize them,the recognition rate in different lighting environments is over 96%.