Dyeing Computer Color Matching Algorithm Based On Improved Bp Neural Network Research

In this paper,due to the shortage of time-consuming and poor precision of the traditional color matching methods,the neural network technology is introduced into the field of textile color matching.Focusing researching on the performance of different kinds of neural networks used in computer color matching for textile dyeing, the model of color matching for textile dyeing with good precision is established by improving the training method of the neural network.Firstly,this paper starts with the analysis of the general theory underlying the BP and RBF neural network,and then establishes the model of textile color matching respectively combining the two kinds of the neural networks.The simulation process is demonstrated and comparisons of simulation error and network performance are discussed.We have analyzed the advantages,disadvantages and improving measures in the paper.Secondly,considering the shortage of poor precision and slow convergence speed of the BP neural network,we make use of LM algorithm and OWO-HWO algorithm to optimize the BP algorithm respectively and the simulating process of the model is implemented.Simulation results show that the two algorithms are more effective than standard BP algorithm either in convergence speed or in training precision.Due to the shortage of easy to trap into local minima,the genetic algorithm is introduced into the BP neural network to improve the network and GA-LMBP algorithm is proposed.In the proposed model,we first use the genetic algorithm to optimize the initial weight values of the BP neural network to locate a better solution space.Then the LMBP algorithm is used in the small solution space to find out the optional solution.This algorithm can overcome the shortage of easy to trap into local optimum of the BP neural network and slow convergence speed of the genetic algorithm with large scale of the group.Finally,aiming at the sample of the dark,medium and light color pigments,we make use of the BP,RBF,OWO-HWO,LMBP and GA-LMBP neural network to make the simulation experiment respectively.Comparisons of simulation error and generalization are also discussed.The simulation results show that the neural network based on OWO-HWO algorithm is suitable to train the sample of light color and the network based on GA-LMBP algorithm is suitable to train the sample of the dark and medium color.