BP神经网络优化Stearns-Noechel模型的羊毛色纺纱配色
BP neural network optimization of Stearns-Noechel model for color matching of wool color spinning yarns
史帅杰 1李启正 1裘柯槟 2朱杰 3张斌 3纪乐福 1陈维国4
作者信息
- 1. 浙江理工大学 纺织科学与工程学院,浙江 杭州 310018
- 2. 嘉兴南湖学院,浙江 嘉兴 314001
- 3. 浙江中鼎纺织科技有限公司,浙江 嘉兴 314511
- 4. 浙江理工大学桐乡研究院有限公司,浙江 嘉兴 314599
- 折叠
摘要
为了提升羊毛色纺纱配色的精确度,通过数理统计方法研究颜色特征中的色相、明度、饱和度与Stearns-Noechel模型参数M值之间的关系,采用BP神经网络对Stearns-Noechel模型参数M值进行优化,并与传统的最优平均M值和波长优化M值等方法进行对比.结果表明:采用BP神经网络优化Stearns-Noechel模型的配色方法比其他 2 种传统优化方法在颜色预测精确度上都有提高.在 99 个羊毛混色纱试验样本中,BP神经网络优化方法得到的平均色差最小,为 1.177 3,其中色差小于 1 的样本占 54%,结合颜色特征采用BP 神经网络优化的Stearns-Noechel模型参数具有较好的效果,对羊毛色纺纱的颜色预测精确度有较大的提高.
Abstract
In order to improve the accuracy of color matching of wool color spinning yarns,the relationship between hue,brightness,saturation in color characteristics,and the parameter M value of the Stearns-Noechel model was investigated by mathematical and statistical methods.The parameter M value of the Stearns-Noechel model was optimized by using the backpropagation(BP)neural network,and compared with the traditional methods such as the optimal average M value and the wavelength-optimized M value.The results show that the optimization of color matching methods of the Stearns-Noechel model using the BP neural network has some improvement in color prediction accuracy than the other two traditional optimization methods.Among 99 samples of wool color-blended yarns,the BP neural network optimization method obtained the smallest average color difference of 1.177 3.In addition,54%of these samples had a color difference of less than 1.This indicates that the use of combining color features and the optimization of parameters of the Stearns-Noechel model using the BP neural network has good results and significantly improves the accuracy of color prediction accuracy of wool color-blended yarns.
关键词
色纺纱/Stearns-Noechel模型/BP神经网络/颜色预测/颜色特征Key words
colored spun yarn/Stearns-Noechel model/backpropagation neural network/color prediction/color features引用本文复制引用
基金项目
中国纺织工业联合会科技指导性项目(2023028)
出版年
2024
NETL
NSTL
维普
万方数据