基于LDA与改进布谷鸟算法的CNN-GRU网络木材染色配方预测  

Wood Stain Formulation Prediction with a CNN-GRU Network Based on LDA and Enhanced Cuckoo Search Algorithm

作　　者：管雪梅[1] 崔宏博 GUAN Xue-mei;CUI Hong-bo(College of Computer and Control Engineering,Northeast Forestry University,Harbin 150040,China)

机构地区：[1]东北林业大学计算机与控制工程学院,哈尔滨150040

出　　处：《科学技术与工程》2024年第28期12268-12276,共9页Science Technology and Engineering

基　　金：国家自然科学基金面上项目(32171691);黑龙江省自然科学基金联合引导项目(LH2020C37);哈尔滨市科技局制造业创新人才项目(CXRC20221110393)。

摘　　要：为了精准预测木材染色配色配方,提出一种线性判别分析(linear discriminant analysis,LDA)和改进布谷鸟算法(improved cuckoo search,ICS)与卷积神经网络(convolutional neural network,CNN)与门控循环单元(gated recurrent unit,GRU)的混合神经网络模型。该模型通过LDA处理光谱信息对其进行分类降维;利用CNN提取重要特征;将这些特征输入GRU中进行训练;网络中的超参数由ICS算法进行寻优。该模型的表现通过多种评估标准进行测量,包括决定系数R^(2)以及国际色差计算公式(CIEDE2000)等。在与多种传统模型的比较中,模型表现出优异的性能。此外,该模型的参数数量相对较少,计算效率高,且稳定性和可靠性良好。结果表明:将该模型应用于通过光谱信息进而预测木材染色配色配方问题上显示出了明显优势。To accurately predict wood dye color matching formulations,a hybrid neural network model combining linear discriminant analysis(LDA),improved cuckoo search(ICS)algorithm,convolutional neural network(CNN),and gated recurrent unit(GRU)was proposed.The model processed and classified the spectral information by LDA for dimensionality reduction,extracted essential features utilizing CNN,input these characteristics into GRU for training,and optimizeed the hyperparameters in the network using the ICS algorithm.The model's performance was measured through various evaluation criteria,including the coefficient of determination R^(2) and the color difference calculation formula(CIEDE2000).In comparison with multiple traditional models,the proposed model demonstrates superior performance.Additionally,the model has a relatively low number of parameters,high computational efficiency,and excellent stability and reliability.The results show that the proposed model exhibits significant advantages when applied to predicting wood dye color matching formulations based on spectral information.

关 键 词：深度学习 木材染色 改进布谷鸟算法(ICS) 

分 类 号：TS664.1[轻工技术与工程]