Comparative Study Of Numerical Analysis And Support Vector Machine In The Dyeing Color

Generally speaking,there are mainly two traditional methods to realize pigments recipe prediction for textile dyeing:spectrophotometric and colorimetric matching.Both of these methods are based on Kubelka-Munk theory.But there are some assumptions in this theory which can cause big errors on calculation.So the recipe obtained is not accurate enough to meet the product demand.The essence of color matching for textile dyeing is a problem of nonlinear modeling with experimental data.As neural network is widely used in nonlinear modeling,many researchers try to use it in color matching and have achieved some result.However, traditional neural network approaches have suffered from difficulties with generalization and local minimum.In order to overcome the deficiencies of the existing color matching methods,data analysis and support vector machine(SVM) thoughts are introduced to color matching processes,which provide two new ways for computer color matching.Computer color matching with numerical analysis method is discussed in this paper. This method establishes the mathematics model based on the mixtures of three colorants. By doing least squares fitting and using Newton-Krylov algorithm,the solution of dyeing recipe has been realized.Experiments prove that this method has showed a prospective result for color matching in textile dyeing which is better than that of BPNN.SVM is a new general learning method based on the statistic learning system which can be used as an effective means to process the non-linear classification and regression.Computer color matching based on SVR algorithm is discussed in this paper.RBF function is chosen as kernel function through Cross-calculation and Particle Swarm Optimization algorithm is used for parameter selection.The model with support vector regression was established and a good result has been achieved.The prediction result demonstrates that SVM is more practicable and effective in the modeling of color matching for Textile Dyeing in comparison with numerical analysis and BPNN methods.The two methods proposed in this paper provide new references for textile dyeing color matching and also have certain values for theory study and application research.