Research On Fabric Defect Detection System Based On Machine Vision

In the textile field,the accuracy of fabric defect detection results is a key factor in judging the quality of fabrics,which carries huge implications on the benefits of textile enterprises.In the actual industrial production,due to machinery breakdowns,the ambient environment and anthropogenic factor,etc.Different types of defects such as stains,yarn breakage,and breakage will inevitably appear.With the fast-growing development of image computer recognition technology,the traditional manual method of detecting fabric defects has gradually been replaced by the method of machine vision technology,and it has a very broad range of applications industrial manufacture.Therefore,this article is based on computer vision and image recognition technology,corresponding to the fabric defect detection theory and algorithm research,using a large number of fabric image samples for testing and experimental results analysis.The research content mainly includes the following aspects:First,the fabric defect detection system is analyzed and researched.In terms of the design requirements,the hardware and software part are designed and developed.Among them,the main body of hardware includes illuminant,camera,lens,etc.,which are respectively selected,position design and analysis calculation;software system includes development environment introduction,system function design,analysis of operation process and development of human-computer interaction interface.Secondly,according to the characteristics and causes of fabric defects,a fabric defect detection algorithm based on machine vision is designed.The algorithm is divided into three parts.The first stage is image preprocessing,including image compression based on discrete cosine transform and median filter noise reduction;the second stage is defect enhancement,the texture background has a great influence on the detection result,a fabric defect enhancement algorithm based on the relative total variation model is proposed to suppress the fabric background texture and extract the edge of the defect;the third stage is defect segmentation,on account of analyzing the gray-scale histogram of the defect-enhanced image,it uses the iterative the adaptive threshold to segment the defect area,and then the image is operated by morphology to form the connected domain to prepare for subsequent defect recognition and evaluation.The test results show that the algorithm can effectively suppress fabric texture and segment defect areas.Finally,for the fabric image after defect detection,this stage extract its geometric and texture features to build a data set,using genetic algorithm(GA)to optimize parameters of the support vector machine(SVM)parameters,and structure GA?SVM multi-classification It realizes the classification task of different grades of fabric defect images for several common defect types,completes the classification and evaluation,and detects 400 different types of fabric images with an accuracy rate of over 94%.After experimental tests,The hardware platform,software system and recognition algorithm of the fabric detection system have reached the expected goal.