Application Of Neural Network Technology In Printing And Dyeing Color Matching

Color of textile has extremely important influence to the product quality, but in many printing and dyeing enterprises the technology of color prescription prediction is heavily dependent on experience of colourist. Color of textile can not be stably remained. The current market demand for the product of textile also becomes to incline to be small batch, fast delivery. These changes have greatly increased the printing and dyeing enterprise workload and cost, so these enterprises urgently need to use more efficient and reliable methods of computer color matching to respond to market changes.Traditional methods of computer color matching are based on single constant Kubelka-Munk theory of tristimulus value matching method. This method needs to make optical database of grey in different concentration gradient before using color prescription prediction. This process takes too long and lack of generality, coupled with the complexity of the relationship between the mixed dye, so traditional color matching method is difficult to meet the current requirements for textile color prescription prediction, printing and dyeing enterprises needs a new simple and efficient matching method.BP neural network has a good nonlinear mapping ability. It can map the complex nonlinear relationship between the color matching prescription and the optical value of the fabric through learning the teacher samples. This paper firstly makes a comparative analysis of the principle and characteristics of the existing computer color matching method, proposed the new using BP neural network method to realize color recipe prediction. Then, it introduces the principle and operation mechanism of BP neural network, and theoretically analysis the advantages and disadvantages. For the shortcoming of BP algorithm easy to fall into local extremum, genetic algorithm is used to optimize it. In the process of building neural network model,this paper introduces color measurement related knowledge, solving the problem of color measurement in computer matching color,and then uses the empirical formula calculate the number of sample teachers. In the last this paper determines the structure of neural network by combining empirical formula and experimental method. The fourth chapter firstly introduces the principle of genetic algorithm and uses the algorithm to determine the neural network initial weights, threshold and learning rate. The method reduces the randomness of the neural network running parameters,and then improves the prediction ability and the convergence speed of neural network. Finally, it talks about the experiment on a variety of neural network. The experimental results show the validity and reliability of the genetic algorithm to optimize the BP neural network in color prescription prediction. This color prescription prediction method solves the problem of making large optical database in traditional computer color prescription prediction, adding a new direction for the method computer color prescription prediction.