Research On Optical Cable Surface Image Acquisition Technology And Defect Recognition Algorithm

Optical cable surface is solidified by black glue,its production process is complex,will go through melting,glue injection,cooling,drying,printing process.Usually with the production equipment performance decline,optical cable surface will inevitably appear damage,scratches,holes and other defects,thereby reducing the reliability of optical cable and transmission performance.China is a major exporter of optical fiber products,if the export of this quality of products,will damage our country's international market competitiveness,therefore,it is necessary to effectively detect the quality of this problem in the production process of optical fiber.The application of machine vision technology in detecting surface defects in optical cable production line to eliminate unqualified products has become an inevitable trend of enterprise production development.To this end,this paper has carried out the following research work.(1)Research on fiber optic cable image acquisition technology.High quality of the original image can decrease the difficulty of algorithm design.Optical image acquisition system mainly includes the lighting unit and image acquisition unit,based on the investigation of the optical image acquisition camera,lens and light source color,light angle and light source control circuit,the design of optical fiber cable image acquisition unit and lighting unit is given to realize the acquisition of high-quality images.(2)Fiber optic cable surface image pretreatment technology optimization.First,the optical fiber diameter detection algorithm is compared,and finally the non-edge method is selected,that is,the image threshold segmentation,and then record the number of pixels in each column of the white optical fiber region and use the average value to calculate the diameter.Then the optical cable image correction algorithm is studied,the optical cable jitter model is established,and the distorted image is compensated by the reverse compensation method.Finally,the image mosaics algorithm is studied,and the image mosaics is realized by using correction,cutting and anisotropic filtering algorithm.(3)Research on feature-based optical fiber surface defect recognition algorithm.Because the optical cable image is a grayscale image,the classification of tape and damage is realized by extracting the aspect ratio,area,circumference and grayscale mean of the defect.By analyzing the shape and gray histogram of the defect of the pinhole,according to its unique shape and the distribution of pixel points with low gray value in the center area,the number of pixel points with low gray value and the gray difference between pixel points are used to identify the pinhole,and the recognition rate is 90%,which meets the user's needs.(4)Deep learning-based optical fiber surface defect recognition algorithm.Increased the defect sample by using the method of defect extraction,paste,and improve the recognition rate of the YOLOv3 model.And compared with the traditional feature-based methods,the detection accuracy of tape,breakage and pinhole is 100%,100% and 80% respectively.The comprehensive recognition accuracy and anti-interference ability of YOLOv3 algorithm are better than the traditional methods.However,when the image has dust interference similar to the epinhole,the false alarm rate of the pinhole is higher,indicating that the recognition rate of the small object is lower than that of the large object.