基于神经网络的人造板装饰纸表面色泽特征分类研究

发布时间：2018-04-03 06:33

本文选题：人造板装饰纸　切入点：表面色泽特征分类　出处：《林业工程学报》2018年01期

【摘要】：通过测定人造板专用装饰纸表面的色泽参数(色度学参数和光泽度),对装饰纸表面色泽特征进行量化分析,并利用色泽参数的特征信息结合误差反向传播神经网络(BP神经网络)对装饰纸进行建模分类,探讨利用装饰纸表面的色泽参数进行装饰纸表面色泽特征分类。以色泽参数数据作为神经网络的输入变量,装饰纸类型作为神经网络的输出变量,建立三层BP神经网络模型,其中,隐含层的最佳节点数为9。结果表明:通过对色泽度参数的主成分分析,增加了光泽度参数后,各类装饰纸之间的独特性增强,更利于对装饰纸进行分类。利用色度学参数(明度指数L*,红绿轴色品指数a*和黄蓝轴色品指数b*)对装饰纸进行建模分类时,判别的总正确率为80.9%,引入光泽度参数之后判别的总正确率提高至92.9%,说明利用色度学与光泽度参数结合BP神经网络可以用于装饰纸表面视觉特征的量化分析以及快速识别分类。
[Abstract]:By measuring the color parameters (chromaticity parameters and glossiness) on the surface of decorative paper for wood-based panels, the color characteristics of decorative paper surface were analyzed quantitatively.Using the characteristic information of color parameters and error back propagation neural network (BP neural network), the paper models and classifies decorative paper, and discusses how to classify the color characteristics of decorative paper surface by using the color parameters of decorative paper surface.Using color parameter data as input variable of neural network and decorative paper type as output variable of neural network, a three-layer BP neural network model is established, in which the best number of nodes in hidden layer is 9.The results showed that by the principal component analysis of color and lustre parameters, the uniqueness of all kinds of decorative paper was enhanced with the addition of gloss parameters, and it was more favorable to classify decorative paper.In the modeling and classification of decorative paper, the colorimetric parameters (brightness index, red and green axis color index a * and yellow and blue axis color product index b) were used to model and classify decorative paper.The total correct rate of discrimination is 80.9, and the total correct rate of discriminating is increased to 92.9 after introducing the gloss parameter. It shows that the combination of chrominance and glossiness parameters combined with BP neural network can be used for the quantitative analysis of the visual features of decorative paper surface and the fast recognition and classification.
【作者单位】：中国林业科学研究院木材工业研究所;
【基金】：国家自然科学基金(31370711) 国家重点研发计划项目(2016YFD0600706)
【分类号】：TS761

【相似文献】