Research On Online Detection System Of Cheese Yarn Defects Based On Machine Vision

Textile industry is closely related to people's daily life,but also as a pillar industry in our national economy occupies a very important position.However,the textile industry is still a labor-intensive industry today,especially in the defect detection process of textile products,which has been relying on manual testing methods seriously restricting the rapid development of the industry.Therefore,it has become one of the urgent problems to find a fast and accurate method to detect textile defects in the production process.Cheese yarn is the product of the last process in the spinning process-winding,whose quality directly affects the production of subsequent processes and ultimately affects the quality of textile products.Due to the quality of yarn raw materials,equipment performance,process flow,personnel operation and other reasons,cheese yarn is prone to produce defects such as multi-layer table,reticulate yarn,chrysanthemum core and multi-source yarn in the forming process.This subject takes the cheese yarn as the research object,chooses industrial computer,light source group,camera set,photoelectric switch and other hardware equipment to set up the detection platform,applies machine vision technology to study the defect detection system suitable for industrial production process field,fast on-line detection of various types of defects is realized.The main research contents are as follows:(1)In view of the uneven profile of the multi-layer defects of the cheese yarn,a high contrast profile image of the cheese yarn is obtained by active light imaging,then the starting and ending points of the side are determined in the contour edge line.Finally,by measuring the error distribution of the profile and the fitting profile,we can measure whether it is a multi-layer defect according to the fitting situation.(2)A defect detection method based on multi-directional matching filter is proposed to overcome the defects of reticulate yarn on the top of cheese yarn.Firstly,the image moment method is used to locate the center of the annular image on the top surface of yarn.Then polar coordinate transformation is used to expand the annular image on the top to a rectangular image.Then Do G algorithm is used to enhance the image edge.Based on the principle of local direction consistency,multi-directional template is used to effectively extract the mesh pixels under strong interference background.Finally,connection relationship is used to further optimize the defect detection results.(3)A fast defect detection method based on image projection is proposed to detect the defects of chrysanthemum core and multi-source yarn.After pretreatment,the two defects have similar image texture features.Firstly,the circular image of the top of the cheese yarn is expanded into a rectangular image according to the geometric correction method of the top of the cheese yarn,and then the expanded image is projected in the vertical direction,by calculating the number of peaks on the projection curve,the chrysanthemum core defects can be fast judged.Projecting the expanded image in the horizontal direction,and by calculating the projection gray deviation value in the range of yarn radius of the cheese yarn,the multi-source yarn defects can be fast judged.(4)An on-line detection system for cheese yarn defects with double exposure mechanism is designed.The images captured under the light source overexposure mode and the normal exposure mode are used for image location and feature extraction respectively.According to the detection index,the selection of key devices is made from the parameters of camera and lens,the type of visual light source and illumination mode,and the type of photoelectric switch,etc.Through the optimization of hardware system workflow,software algorithm detection process and man-machine interface design,the automatic on-line detection of cheese yarn is realized.