Research On Color Difference Online Detection System Based On Image Processing In Fabric

Textile manufacturing is one of the major global industries.The textile industry develops rapidly and competition is increasingly fierce especially in China,which is a large country of textile clothing production and export.Many textile enterprises are suffering unprecedented pressure at the same time.In order to keep long-term foothold in the textile industry,Enterprises must improve the technology innovation,advanced the equipment,improve the service industry chain and reduce the cost.At this stage,however,the paucity of research on the textile color difference detection has made many small and medium-sized enterprises still rely on artificial visual method to detect textile color difference,this kind of detection method exists many deficiencies: First,person's subjective evaluation is easy to cause misjudgment,especially after long work visual fatigue;Secondly,this method is time-consuming,severely reduces the competitiveness of the enterprises.Finally,will cause the waste of resources and improve the production cost.To solve these problems,this article based on machine vision and image processing technology to realize intelligent textile color detection and classification of color fabrics.This topic researches and analyzes the current situation at home and abroad firstly.In terms of system hardware structure,D65 standard light source and high sensitive CCD industrial camera are chosen.In order to reduce the effects of noise on image processing process,and guarantee the integrity of the image color information,the Gauss filter is used for image preprocessing.Then the related theory of color like color space model and color difference formula are expounded.Secondly,this paper introduces the related theory of fabric color similarity algorithm,and puts forward a kind of partitioned off color detection algorithm based on T-S fuzzy neural network.A large number of textile samples images are collected,and RGB values of images are extracted.And then samples are trained by T-S fuzzy neural network to realize the converting from RGB to the LAB space,which make the LAB color extraction process is more convenient and quick,as well as avoid the massive calculation of traditional color space conversion formula.Through calculating color difference of the fabric image with blocking,the position of the color abnormal point can be located in time.Making the experimental results and the results measured by SP60 spectrophotometer into comparison,the accuracy of the algorithm is verified.At the same time,a simple visual color difference detection system interface is designed to convenient staff understand the detailed results of the inspection in real-time.In terms of textile color automatic classification,this paper compared three classification algorithms: the standard SVM,the LS-SVM and SOM neural network and puts them into the color classification application of textiles.Through the experiments,taking classification accuracy and efficiency into consideration,LS-SVM can better realize online classification of textile color.In this paper,the color difference of monochromatic textiles automatic on-line detection and color classification are studied,which can be exploited to realize color quality automatic controlling of the textile industrial production.