摘要
针对传统配色方法及配色算法存在不足之处,利用BP神经网络对色纺纱进行配方预测,并用遗传算法对其进行改进.结果表明:将遗传算法引入到BP神经网络,可优化BP神经网络配色模型;测试样本包含在训练样本中时,预测配方精度非常高,配方绝对误差均值几乎为0;而测试样本不包含在训练样本中时,预测配方精度较低,配方绝对误差均值为0.033,初次打样色差均值为1.69 CMC(2∶1),大于1 CMC(2∶1).
For the deficiencies of traditional color matching and color matching algorithm, BP neural network is used to pre-dict the formula of the colored spun yarn, and the genetic algorithm is introduced to improve the BP neural net-work. The results show that ： BP neural network can be optimized when the genetic algorithm is introduced into it , but when the test sample is contained in the training sample data, the color matching accuracy of this Ga-BP neural network is very high and the mean formula absolute error is almost 0 , while when the test sample is not in-cluded in the training samples, the color matching accuracy is lower and the mean formula absolute error is 0.033 , the mean color difference of the first smaple is 1.69 CMC（2 ： 1 ） , and more than 1 CMC（2 ： 1）.
出处
《天津工业大学学报》
CAS
北大核心
2016年第6期27-31,共5页
Journal of Tiangong University
基金
国家重点研发计划专题(2016YFB0302801-03)
关键词
色纺纱
BP神经网络
遗传算法
配色
colored spun yarn
BP neural network
genetic algorithm
color matching
