Grade Evaluation Of Color Fastness To Laundeirng Based On Image Analysis

During the inspection of textiles for import and export trade, the color fastness tolaundering of textiles is mainly depended by the dye qulity to a larger content.In the current international and national standards, grade evaluation of colorfastness to laundering is still evaluated visually according to the reference sample byexpert. But visual evaluation is a subjective method, the effectiveness and consistencyof the result depends on the observer.In order to evaluate the color fastness grade after frequent laundering objectively,four methods based on image analysis is proposed. And then the feasibility of thecomputer vision evaluation for color fastness is explored in this paper.In the first method, firstly the replicas of different grades of the Gray Scale forColor Change and Gray Scale for Staining for AATCC61-2010are scanned tocalculate the gray value. Secondly, the equation of grade for color change-gray-scalevalue and the equation of grade for staining-gray-scale value are fitted by analyzingthe relationship between the gray value of color change and its corresponding grade.Finally, the grayscale average of samples after the laundering are calculated beforesubstituted them into the fastness rating equation. When compared with the visualevaluation, the method based on color fastness grade equation is much simpler andobjective.In the second method, the weighting method is used to analyze the grade of colorfastness of the samples for color chang. Firstly, the grey value of discolored area isarranged in ascending order. And then, the corresponding weights of9levels arecalculated through the gray scale for color change. Finally, the weighted grey valuesof samples are substituted into the equation of grade for color change-gray-scale valueto compute the grade of color change of the sample.In the third method, the staining grade evaluation based on Fourier transform isproposed to eliminate the interference of fabric texture. In this method, a sampleimage without texture information is gotten after Fourier fitering with the unevenstaining sample. Finally, the grey values of samples without texture information aresubstituted into the equation of grade for color change-gray-scale value to computethe grade of color change of the sample.In the forth method, the staining grade evaluation based on waveletr transform isalso proposed to eliminate the interference of fabric texture. In this method, thelow-frequency sub-graph of the staining sample reconstructed after waveletdecomposition for4layer. The reconstructed image complete retained the sample color information. Finally, the grey values of samples are substituted into the equationof grade for color change-gray-scale value to compute the grade of color change ofthe sample.To verify the reliability of two methods for color change grade evaluation, thegrades of18samples are evaluated by the first and second methods are compared withthe results of visual rating. The experimental results indicate that there are16data thatare consistent with the visual rating method. Only seven data of the results based onequation method are consistent with visual data because of the nonuniform colorchange of samples. When verify the reliability of three methods for grade evaluationafter staining, the48experimental results indicate that the accuracy of equationmethod is low, the results of Fourier transform method are much closer to the visualgrating method.While the evaluation results based wavelet tranfrorm indicate that the41samples out of the total48test samples are consistent with the visuall ratingmethod, and the wavelet transform method has no special uniformity requirement ofthe sample, which can be widely applied.