Research On Application Of BP Neural Networks In Computer Color Matching For Textile Dyeing Based On Hidden Layer Improvement

Recently,the theory of textile dyeing and color matching are developed very quickly, and the computer technology becomes more and more popular.This has made it necessary for the field of textile color matching to combine the theory with the computer technology. Due to the shortage of traditional color matching methods applied with BP neural network, an improved algorithm is put forward in this paper.The new algorithm references the outputs of hidden layer.In this algorithm,the input of hidden layer combines the input date and the output of hidden layer.It focues establishing a model of color matching for textile dyeing with good precision and higher convergence speed.This paper analyses the theory of textile dyeing and color matching and the causation of chromatism from the aspect of CIE chromatic colorimetry.The traditional computer color matching methods are based on the Kubelka-Munk theory.On the basis of the theory,tristimulus values and reflection spectrum methods are advanced.Although these two typical methods are efficient,they are not applied directly for textile dyeing and color matching for theirs shortages.The former is only adapt to resolve the problem about only three dyeing consistence'values,the latter is inefficacy to deal with the non-linear relation between the K/S value and the chroma.The Artificial Neural Networks has its predominance's such as,it is a self-adaptive method,its concurrent distributed memory mode.So the application of the ANN spreads widely.This paper choices BP Neural Networks from the models of ANN to apply on the textile dyeing and color matching.After training the BP Neural Networks by the date propagation in positive direction and the error values propagation in reverse direction,a suitable network is build up.Although the BP network has strong non-linear advantage,it has many disadvantages such as the tendency to becoming trapped into local minima,poor precision and slow convergence speed.Then a network is designed and trained in this paper by a modified training algorithm based on BP neural network,which references the output of hidden layer,for textile dyeing computer color matching.This network is simulated and implemented with MATLAB.We make use of the traditional and new BP neural network to make the simulation experiment respectively aiming at the sample of the dark,medium and light color pigments.The possibility and reliability of the modified BP network used in color matching for textile dyeing have been confirmed by experiments.