Research On Image Feature Analysis And Detection Method Of Weft Knitted Fabric Defects

Textile fabric defects are of various types and shapes.In terms of the morphological features of the defect image,there are abnormal gray level features and abnormal texture features.According to the weaving processing method,usually can be divided into two major categories of woven fabrics and knitted fabrics,textile fabrics compared with woven fabric,knitted fabric,especially the mechanism of weft knitted fabric weaving is more complex,more diverse organization texture structure,therefore,easier to produce in the process of weaving texture features changes but gray level change is not obvious defect,according to this kind of defect detection problem,There are shortcomings in the existing methods.Fabric defect detection method research at present,mainly for woven fabric,knitted fabric as the object of the research content is less,and lack of fabric defect image morphology research,not to quantify the exception class texture defect shape characteristic description,resulting in abnormal for gray and texture abnormal defect,failed to put forward a clear defect theory and method of image classification,in addition,The specific quantitative relationship between the extraction method of fabric defect feature and the morphological feature of defect image could not be established.As a result,the detection rate of existing fabric defect detection method on knitted fabric defect,especially the abnormal texture defect of knitted fabric,was low.The main knitting defects of weft knitted fabric and their causes were studied by analyzing the weaving mechanism,structural characteristics and influencing factors of weft knitted fabric.From the perspective of graphics,it was concluded that the weft knitted fabric defect image had two forms of abnormal changes of gray distribution and abnormal changes of texture mode.A method of approximate description of defect image features by abnormal variation of vibration characteristics of two dimensional vibration signals is proposed.Based on the expression of the defect image and the gradient feature of the edge of the defect contour,a classification theory of the defect image was proposed.The defect image was divided into grayscale pattern defect image and texture pattern defect image,and the classification operator of the defect image was constructed.Firstly,the texture gradient peak value of the normal non-defect weft knitted fabric image_-was calculated.Then,the mean value of gradient probability of defect image to be detected_-80))is calculated.Finally,the defect images were classified by comparing_-and_-80)).The parameter setting problem of classification operator is discussed and its validity is verified by experiments.Aiming at the grayscale morphological defect image,based on the regional characteristics of image gradient distribution,the improved PM model with 8 diffusion directions and 3 groups of diffusion coefficient functions was introduced to carry out nonlinear diffusion processing on the defect image,and the optimization operator of setting gradient threshold₁and₂was constructed to realize feature extraction of grayscale morphological defect image.The grayscale defect detection channel was constructed,and the non-linear diffusion image was threshed by OTSU threshold segmentation method to detect grayscale defect.Aiming at the texture morphological defect image,firstly,based on the image scale spatial distribution characteristics,the WEMD algorithm was introduced to decompose the defect image adaptively,and the component image of texture feature information was obtained.Then,a combination feature extractor was constructed to extract texture feature information from fabric texture feature information component images.The feature information fusion technology was used to eliminate the feature information with low correlation and redundancy,and the texture combination feature vector was formed to realize texture morphological defect image feature extraction.The texture defect detection channel was constructed,and the texture combination feature vector was substituted into the DE-HHO-RELM.On the basis of traditional weft knitted fabric inspection equipment,the improvement scheme of fabric defect detection system based on machine vision was studied,and the selection of industrial camera,light source equipment,image acquisition and processing equipment was discussed.Based on the analysis of biological vision defect detection steps and judgment basis,a double channel weft knitted fabric defect detection algorithm was proposed.The algorithm firstly used the defect image classification operator to judge the possible defect types in the fabric image to be detected,then input the image to be detected into the grayscale pattern defect detection channel or texture pattern defect detection channel according to the classification result of the defect image,and finally output the defect detection result.The system test results show that the weft knitted fabric defect detection system designed in this paper has higher detection rate and lower error detection rate.The existing researches are lack of the research content on the image features and detection methods of knitted fabric texture defects.Therefore,this paper took weft knitted fabric as an example to analyze the image features of knitted fabric defects,and discussed the targeted detection methods for grayscale and texture defects.In this paper,the morphological characteristics of weft knitted fabric image were studied,the classification theory and operator of defect image were constructed,the detection channel of grayscale morphological defect and texture morphological defect were constructed,and the detection system of weft knitted fabric defect based on double channel defect detection algorithm was designed and realized.The research results of this paper are currently being applied in weaving enterprises in the form of products,which is of great significance to the optimization of weft knitted fabric defect detection theory and method,and lays a foundation for the subsequent analysis and detection of jacquard fabric texture anomaly variation defect characteristics with more complex texture pattern.