Research On Appearance Detection System Of Bobbin Yarn Based On Machine Vision

The bobbin yarn is a tubular yarn body formed by winding electronic glass fiber on the bobbin.Due to the influence of the production environment and the reuse of the bobbin,it is necessary to pass through the appearance quality detection process before putting into use.It mainly includes hairiness detection,surface defect detection and bobbin defect detection.According to enterprise research,most glass fiber enterprises still rely on manual inspection of the appearance of the bobbin yarn,although this method has disadvantages of low detection efficiency and high labor intensity.This thesis designs the bobbin yarn appearance detection system to realize the automation of the bobbin yarn appearance detection based on machine vision technology.The main work is as follows:(1)This thesis analyzes the research status of machine vision detection technology and bobbin yarn appearance detection at home and abroad,introduces the overall design scheme of the bobbin yarn appearance detection system,designs the machine vision hardware platform and selects the required vision hardware.(2)Bobbin yarn hairiness detection algorithm.The bobbin yarn hairiness detection is divided into three steps: hairiness extraction,hairiness classification,and hairiness statistics.Firstly: border following,Hough transform and other algorithms are used to extract hairiness.Secondly: The moment features and region features are selected and hairiness classification is realized by support vector machine.The length of the hairiness is measured by the principal component analysis algorithm and the minimum outer rectangle method and the number of hairiness is counted at the same time.Finally: The effectiveness of hairiness detection algorithm is verified by comparing with the results of manual detection.(3)Bobbin defect detection algorithm.Firstly: The image of the bobbin is enhanced by a specific high-pass filter.Then the Canny algorithm is used to extract image edges and the detection ROI(Region of Interest)is obtained according to the image edge features.Finally,frequency-tuned salient analysis algorithm is used for defect segmentation.The experimental results show that the combination of ROI extraction algorithm based on edge features and frequency-tuned salient analysis algorithm can achieve the bobbin defects segmentation accurately.(4)Bobbin yarn surface defect detection algorithm.The image of the bobbin yarn surface is enhanced,standard multi-scale template is established and location of bobbin yarn surface is realized by the normalized correlation coefficient template matching method and acceleration based on integral map.Then the defect classification is realized by the residual network(Res Net50),and compared with the classification result of using the support vector machine and gray level co-occurrence matrix.The experiment shows that the classification accuracy of bobbin yarn surface image by using Res Net50 can reach99.17%.(5)Finally,the system's human-computer interaction interface is designed,and the overall system test is carried out.The results show that the system can detect the appearance of the bobbin yarn accurately and efficiently,and achieve the automatic screening of qualified and data statistics of the bobbin yarn.There are 50 figures,13 tables and 61 references in this thesis.