Design Of Fabric Defect Detection Algorithm Based On Low Rank Decomposition

To promote the intelligent information construction of the textile industry is not only the national strategy but also the industry demand,in which the intelligent automatic detection of fabric defects is the most important and occupies an important position.As an important part of the quality control in textile industry,it directly affects product quality and relates to enterprise income and national GDP.Traditional manual detection cannot maintain the same strict standard for a long time because of the differences of physical condition,subjectivity,skill proficiency and other aspects of workers.Therefore,it is necessary to replace traditional manual testing with objective mechanized automatic detection equipment.Due to the progress of society and the upgrading of production technology,there are numerous fabrics on the market at present.According to the texture difference of the fabric,it can be divided into three categories: plain twill fabric,periodic texture fabric,irregular printing texture fabric.Plain twill fabric is relatively simple and has no pattern on its surface.There are many related research literatures on this kind of fabric because of its early appearance.However,the periodic and the irregular printed textured fabric are the focus of the research,especially the irregular printed textured fabric.There are few literatures on the defect detection of irregular printed fabrics,because such fabrics do not appear for a long time.But this kind of fabric is loved by consumers,and the production has increased rapidly.It is urgent to design corresponding automatic defect detection algorithm.Therefore,this dissertation designs a kind of automatic defect detection algorithms based on low-rank decomposition for periodic and irregular printed fabrics.The main contributions of this dissertation are as follows:(1)An improved low-rank decomposition detection method based on Beta norm is proposed to deal with the problem of excessive loss of image information and the skew interference caused by traditional low-rank decomposition and periodic fabric elasticity respectively.Firstly,a prior map is constructed by extracting the texton feature of the fabric image.Secondly,a Beta norm is used to replace the nuclear norm in the low-rank decomposition method,and the low-rank decomposition is guided by prior map to decompose the fabric image,which solves the problem of excessive loss of image information caused by nuclear norm in traditional low rank decomposition.Furthermore,a posterior map is constructed by extracting the HOG(Histogram of Oriented Gradients)feature of the fabric image,and a saliency map is obtained by Hadamard product between the posterior map and the sparse component obtained by the low-rank decomposition which can solve the skew problem caused by fabric elasticity.Finally,an optimal threshold segmentation is used to obtain the defect figure.Compared with the existing four methods,the experimental results demonstrate that the proposed method can effectively suppress the skew interference and the detection time is shorter.(2)A double sparse low-rank decomposition method is proposed to defect detection for complex irregular printed fabrics.Firstly,the low rank decomposition model with double sparse characteristics is established by taking the sparse component obtained as the printing template prior,and taking the result of the difference between the defective printing fabric map and the template fabric map as the defect prior.Secondly,according to the double sparse low-rank decomposition model of irregular printed fabrics,the decomposition is guided by the prior of printing and defect to obtain the saliency map of defects.Finally,the defect map is obtained by binarization of the saliency map of the defect using the optimal threshold segmentation.The simulation results show that the proposed method is effective and accurate,and the detection time is 10.22% less than the current optimal PN-RPCA algorithm.(3)The software system of fabric defect detection is designed.In this system,three algorithms are used to detect the fabric defects according to the characteristics of plain twill fabric,periodic fabric and irregular printed fabric from the aspects of pertinence and efficiency,and the detection function of algorithm is integrated into the software system.At the same time,the function of the system is verified by the fabric samples collected in the factory,and the results are saved to the corresponding documents.Finally,this dissertation is summarized,and future research directions are prospected based on existing problems and development trends.