Research Of Optical Cable Defect Inspection System Based On Machine Vision

Aiming at the problem of long time existing defects in the cable industry,this paper selects the optical cable as the research object.It carried out the research on the appearance of optical cable defects inspection system based on machine vision cable to achieve real-time detection of cable production.The specific content of my work are:(1)Determine the hardware and systems.In view of the problem of linear motion and fast speed of optical cable production,the linear array camera is used for image acquisition.we calculated the corresponding lens that meet the requirements.We need to select the light source,some accessories such as encoder,acquisition card,the light source controller and so on.At last we determine the detection scheme.(2)Detect the appearance of defects in sheath.For the detection of defects and the diameter of the appearance of concave dent,it is easy to detect scratches and pits of cable diameter by using threshold segmentation method that can distinguish between pits scratches,cable and background.(3)Image mosaic.Because of the long range of characters,the middle containing most of the blank area and the image was too broad to show,we use simple image mosaic to multiple small image combination that contains the character in a cycle.According to the characteristics of the whole character,it can be very accurate and stable splicing out the integrity of the image in the condition of the outside not giving the signal.(4)Character quality detection.The template matching method is used to detect the fixed part of the character which is the appearance of the printing quality detection for cable machinery.We need to create a template and compare the real-time image.It is judged by the number of characters and the similarity of the matching.(5)Digital meter recognition.In view of the change in the number of meters in mechanical printing and the characteristics of the printing,we use neural network classifier to classify digital meter.The recognition rate can meet the requirements of the customer when the printed number is clear and complete.We judge whether the printing wheel printed is jumping code by contrasting the digital real-time with the actual length of the recognition.The test system is installed in the field of practical application.Through a trial run for more than 2 months of continuous improvement,it meets the customer testing requirements and verifies the feasibility and effectiveness of the detection system.The detection system in this paper is not only limited to the detection of defects in optical fiber cable,but also can be used in other similar cable products.It has a good market prospect.