Yarn Appearance Quality Inspection System Based On Machine Vision

With the continuous improvement of living standards,people also put forward higher requirements for the quality of textiles.Yarn as a textile raw material,the evenness of the finished product will affect the appearance quality of the textile,and the defect will cause the end breaking in the weaving process and affect the textile efficiency,so it is necessary to detect and control the evenness and defect of the yarn.The existing detection method of yarn strip and defect is a combination of capacitance method and photoelectric method,which is susceptible to the influence of air humidity and yarn material characteristics,resulting in inaccurate uniformity of yarn(CV%)and defect data.In order to make up for the shortcomings of traditional detection methods,machine vision has been applied to yarn quality detection.However,there are still problems such as poor yarn image preprocessing effect and low yarn segmentation efficiency.There are many isolated points on both sides of the yarn trunk after yarn segmentation or the trunk is damaged,so that the diameter measurement error is large.In view of the above problems,this topic to count 7tex,10 tex,11tex,63 tex,84tex cotton yarn as the research object,the main research content is as follows:(1)Design and build the yarn quality inspection system platform.The system includes image acquisition device,yarn transmission and collection device,image processing and data analysis software.(2)In order to obtain a clear yarn image and reduce the running time and error of the subsequent yarn segmentation algorithm,an improved nonlinear diffusion model(P-M model)is proposed to smooth and remove the noise in the main feature region of the yarn.In general,the P-M diffusion model requires a large number of iterations for noise removal and edge enhancement of the image,so the efficiency is low.In this thesis,the diffusion factor in P-M model is improved by reverse smoothing yarn features,and the traditional method of front and rear item difference is replaced by the method of finding the center difference,so that the detection efficiency and robustness of the algorithm are improved.Finally,the yarn image after denoising and increasing the contrast is obtained.(3)In order to obtain the yarn image without outliers and damaging the backbone of the yarn,a yarn yarn extraction algorithm based on the maximum inter-class variance(OSTU)and template matching is proposed.Firstly,three 4×2 pixel,5×2 pixel and 6×2 pixel regions are selected by rectangular ROI on the yarn image segmented by OSTU segmentation method as template images.In order to ensure the accuracy and applicability of template parameters,ROI of 3×1 pixels was selected as the monitoring template.Finally,yarn images to be matched were extracted for template matching.In this thesis,the method is verified experimentally,and the image of the main yarn with no outlier and no damage is obtained.(4)A measurement method of yarn diameter and a classification and statistical method of yarn defects are designed.According to the yarn diameter data obtained,the yarn evenness and the number of various types of defects are calculated and compared with the measurement results of USTER CLASSIMAT 5.Both the change of yarn evenness and the change trend of the number of various types of defects have a great correlation.Finally,in order to facilitate users to use and operate the detection system,a yarn detection operation software is designed,which integrates yarn collection,preprocessing,camera calibration,detection algorithm,measurement algorithm,data storage,data communication and other functions.There are 94 figures,14 tables and 64 references.