Research On Color Matching In Textile Dyeing Based On RBFNN And FIS

This paper first starts with the analysis of significance and the methods of textiledyeing and color matching, then presents the general theory underlying the RBFnetwork and fuzzy inference system. In addition, the two methods of color recipeprediction based on RBF neural network and ANFIS respectively are studied.In the field of textile dyeing industry, the theory, method and applied research onautomatic color matching always are the important issues. The textile dyeing andcolor matching are involved in massive complex, fuzzy and highly-related variablefactors, therefore its modeling method is always the difficult point of the fundamentalresearch and the application. At present, it is still very difficultly to construct accuratemodels mathematically. The traditional color matching methods depend upon person'sexperience, so it is time-consuming and imprecise. The current computer colormatching (CCM) method is one kind of new technology that combines the moderncolorimetric theory and the computer technology, gain optics information of the colorsamples through the optical instrument, then produces the best color recipe by thecomputer. This method is based on the K-M theory, its precision is high insufficientlyand the result also needs to revise. Due to the unavoidable influence of variouscomplicated factors and textile samples' data fuzzy characteristics during thecoloration process, the tradition and the current computer color matching methods areconfined to some limitations. In consideration of these limitations, based on fuzzyinterference system and artificial neural networks theories which are developed inrecent years, the architecture, principles and related theory of ANFIS which integratesthe best features of fuzzy system and neural networks are studied and its application intextile dyeing and color matching is also researched, which has certain practicalsignificance to further study and application of textile dyeing and color matching.A new color matching model based on ANFIS is proposed in the paper, theproposed fuzzy modeling method integrates fuzzy inference system into artificialneural networks in order to construct ANFIS. Under the platform of NeuroSolutions, the simulation process are demonstrated and comparisons of simulation results arediscussed and listed, and it show that model based on ANFIS is more effective thanthe one based on RBFNN.These two methods provide the new ways for computer color matching andsimultaneously present new references for further application of theory into textiledyeing and color matching, bear the certain fundamental research value and thepractical application value, which is of certain reference values of theoretic study andpractical application.