Research On Application Of BP Neural Network Based On Genetic Algorithm For Color Matching In Textile Dyeing

The process of textile dyeing is as follows: first, the textile sample must be got fromcustomers; second, color matching workers must draw a design according to the color ofthe textile sample. So it is very deficient. Aiming to improve the dyeing efficiency, manytextile dyeing technologies have been proposed.The traditional methods are based on the Kubelka-Munk theory, which refers to theopaque materials' absorbability and reflection. Park and Steams put forward the colorblending formula which is the mathematical foundation of computer color matching.Along with the improvement of the artificial intelligence, the application of thetheory spreads widely. BP is a strong learning algorithm. According to the viewpoint ofinformation disposition, BP network is a kind of information mapping system. The BPneural network can finish any special mapping after supervised learning. So it is the toolof nonlinear system modeling, simulating, and forecasting. Due to these advantages, it isused into textile dyeing computer matching. In this system, we let CMY value be theinput of the neural network, and take the dyeing consistence as the output of the system.In order to improve the generalizing ability of BP neural network, we use a hiding layer.So we adopt a three-layer neural network at last.The BP neural network has the tendency to become trapped into local minima and isvery sensitive to initial training values. Therefore the genetic algorithm is introduced intothe system to improve the system. In the proposed model, we first use the geneticalgorithm to optimize the initial weight values of the BP neural network, and locate abetter solution space. Then the BP algorithm is used in the small solution space to find outthe optional solution.