Study On Defect Detection System Of Patterned Carpet Based On Machine Vision

The textile industry is a kind of important fabrics,its quality detection is still mainly artificial detection.It will lead to the low productivity and high missed rate.Because of the unique characteristics of the vision detection,it is an effective method to detect the defect of the carpet using machine vision technology.Now the researches are mainly focused on the pure color carpet,but patterned carpets are more widely used in the daily life,and the defect detection researches are less.For this reason,a defect detection system of patterned carpet based on machine vision is designed in this paper.This system can effectively detect three kinds of defects,respectively skipped stitches,broken yarn and stains,and it has high detection accuracy and stable algorithm.This paper presents a new method of camera calibration,which has many advantages.For example,the calibration is flexible,the target is simple,the operation is simple and it can be applied to the production environment of the carpet with narrow working space.The patterned carpet local image which is similar to the template image is obtained by the cyclic matching method of the normalized product;the classification of multiple local images obtained by the SVM method,then the local image with suspected defect will be obtained and the categories of suspected defects will be known.Thus the stability,efficiency and the accuracy of the detection algorithm of the defective patterned carpet can be effectively improved.The template image is divided into a brighter image and a darker image,as the template image and the patterned carpet image with suspected defects are almost identical in size and direction,the patterned carpet image will be divided into brighter part and darker part.Then the subtraction is used to extract the defects of the brighter and darker image,if the defects are not extracted,the image is normal.At the same time,some experiments on the patterned carpet have been done.The results show that the vision detection system which is designed in this paper can quickly and accurately extract the skipped stitches,broken yarn and stains in the patterned carpet,which can verify the superiority of the visual detection system designed in this paper.