Design Of Yarn Packages Appearance Defect Detection System Based On Machine Vision

Yarn package is a kind of packaging form for the carriage of chemical fiber.It is produced by four steps: winding,peeling,inspection and packing.The first three steps will cause hairiness,depression and oil stain on the surface of the yarn packages.At present,the defects of yarn packages appearance are detected manually domestic and overseas.In addition,in order to distinguish batches of yarn packages,it is necessary to manually classify colors of paper tubes.Since the yarn packages defects are small,however the detection speed is fast during the process of detection,the stability and reliability of the manual detection are difficult to meet the requirement,which can even cause secondary damage to the yarn packages.In order to realize the automatic detection of yarn packages appearance defects and paper tube color,a fast and stable detection system of the appearance defect of the yarn packages is designed in this paper.The main work is as follows:(1)Summarize and analyze the current research status of appearance defect detection methods and machine vision technology commonly domestic and overseas.According to the defect characteristics of the yarn packages and the detection speed requirements,a hardware platform is built and the model of each component is chosen.(2)Detection and classification of yarn packages hairiness based on unidirectional convex hull and SVM.The morphology and structural characteristics of hairiness are analyzed.Firstly,the hairiness image features are extracted by using a specific convolution kernel.Then the contour detection method and the proposed unidirectional convex hull detection algorithm are used to locate and count the hairiness.Finally,the local binary patterns and SVM are used to classify the straight hairiness and the curl hairiness.And the effectiveness of the method is verified by experiments.(3)Oil stain and depression detection on the surface of the yarn packages based on DoG and coordinate sequences.The oil stain and depression features on the surface of the yarn packages are analyzed,and the oil stain on the surface of the yarn packages and the depression in the upper and lower surfaces are detected using the DoG algorithm.Due to the reflective characteristics of the side of the yarn packages,a special lighting method is selected,and the side depression detection method based on the coordinate sequence analysis is designed.The experimental results show that the Dog and coordinate sequence analysis methods can be used to detect the oil stain and depression on the surface of the yarn packages accurately and efficiently.(4)Color identification of paper tube based on MLP.For the paper tubes of 8 different colors in practical production,the R,G,and B channel average value of the images are counted as paper tube color feature vectors,and a specific multi-layer perceptron is constructed to train them to correctly identify different colors of paper tube.(5)Pycharm is used as the development environment.OpenCV is used as the image processing tool for the appearance of the yarn packages.The detection result is managed and saved by the PySQLite database.There are integrated with the human-computer interaction interface written by PyQt.Finally,the yarn packages appearance defect detection system is completed based on machine vision.In this paper,there are 48 pictures,7 tables,and 79 references.