Research On Color Difference Detection System In Fabric

China is a big textile country.With the implementation and promotion of the “Belt and Road” national strategy,domestic textile manufacturers successfully take advantage of textile exports.However,customers tend to be very strict with textile quality in the international trade of textiles.Color difference is one of the important indicators in the evaluation of textile quality.Many textile printing and dyeing companies do not have qualified on-line color difference detection equipment,so the products are often returned and reproduced when they are delivered.Not only does this cause serious economic losses to enterprises,but also aggravates environmental the pollution and the resources waste.Therefore,it is of great significance to research and design the on-line color difference detection system for textiles,which can avoid the problems of human eye detection with subjective and low accuracy,and off-line colorimetric equipment with limited detection range and low efficiency.In this paper,two different methods are applied to the color difference detection of textiles.First,a method based on machine vision and image processing is adopted.A super pixel image segmentation method is applied to the standard sample image segment,and the segmented path is applied to the sample image being detected,meanwhile the color difference detection between the two regions is performed one by one in the super pixel region.This method can make relatively balance of the detection speed and calculation accuracy,and meet the need for online inspection of fabric color differences.Secondly,a color difference detection method based on spectral data is adopted,too.The Optical fiber spectrometers are used to collect spectral reflectance data for different color fabrics and extract relevant features.The L*a*b* values of each fabric sample are measured and recorded.After the establishment of the PCA-ELM mapping model,a large number of samples are used to train the network,and the spectral reflectance characteristic data and the mapping relationship between L*a*b* values are obtained,instead of the traditional object color tri-stimulus XYZ obtained from the spectral reflectance.The color tri-stimulus XYZ is used to obtain the color physical quantity of L*,a*,b*.After using this model to obtain the value of L*a*b*,the color difference of the sample is calculated by using the CMC(l:c)color difference formula.In the color classification of textiles,Convolution Neural Networks(Alexnet)are used to achieve automatic classification of the different-color textiles.Through simulation experiments,the classification accuracy rate reaches 100%,which is higher than the traditional classification algorithm SVM.The paper focus on the study of the key technologies of online color difference detection and color classification of textiles,and it has a reference value for realizing the automatic control of textile color quality.