A Dyeing Color Matching Method Combining RBF Neural Networks with Genetic Algorithms

This paper makes use of the non-linear relation between dye concentration and textile reflectivity, and the non-linear relation between textile reflectivity and the Lab value of textile to achieve a conclusion that there is a non-linear relation between dye concentration and the Lab value of textile. In addition, this paper uses the RBF (radial basis function) neural network to approximate the relation. The dye concentration is the neural network's input value and the Lab value of textile is the output. To find the optimal formula of color matching, the Genetic Algorithm is also involved in the model. Specifically, let the dye concentration be the individual. The individual serves as the RBF's input, and we use the color difference Delta E between the RBF's output and object color be the fitness function. Compared to the color matching concerning Kubelka-Munk theory, the model possesses some advantages as following: low cost, no assumptions in it, high precision, fast-achieved formula, etc. Experiments prove that this model can quickly provide the users with a satisfying formula.