基于SOA-BP神经网络的显示器颜色特性化研究  

Research on Display Color Characterization Based on SOA-BP Neural Network

作　　者：王荣欣 郭凌华 陈睿 赵甜甜 孙英 WANG Rong-xin;GUO Ling-hua;CHEN Rui;ZHAO Tian-tian;SUN Ying(School of Light Industry Science and Engineering,Shaanxi University of Science and Technology,Xi’an 710021,China;Shaanxi Fenghua Culture Development Co.,Ltd.,Xi’an 710062,China)

机构地区：[1]陕西科技大学轻工科学与工程学院,西安710021 [2]陕西风华文化发展有限公司,西安710062

出　　处：《印刷与数字媒体技术研究》2025年第2期20-28,共9页Printing and Digital Media Technology Study

基　　金：国家自然科学基金项目(No.22078186);陕西省重点研发计划项目(No.2020GY-320)。

摘　　要：为了保证显示设备的颜色显示准确性,在对显示设备进行色彩管理时,需要提高显示器颜色特性化精度。本研究提出一种基于海鸥优化算法(Seagull Optimization Algorithm,SOA)优化BP神经网络(SOA-BP)的显示器颜色特性化模型。首先,采用K折交叉验证方法确定BP神经网络最佳隐藏层神经元个数,建立显示器输入信号RGB值与显示颜色L^(*)a^(*)b^(*)之间转换的BP神经网络模型;然后,通过海鸥算法对BP神经网络的权值阈值进行优化,建立基于SOA优化BP神经网络的显示器颜色特性化模型;最后,对优化前后的神经网络模型进行训练与测试,并进行模型精度对比分析。在仿真实验中,海鸥算法优化BP神经网络预测模型测试10次得到的CIELAB色差△E_(ab)^(*)和CIE2000色差ΔE_(00)平均值分别为2.291和1.032,较优化前分别减少了31.79%和36.18%,且优化后的SOA-BP模型更稳定,说明本研究所建立的SOA-BP网络模型对显示器颜色特性化具有较高的预测精度和较好的稳定性,为印刷包装领域色彩管理颜色特性化提供了理论和实践的参考。In order to ensure the accuracy of the color of the display,it is necessary to improve the precision of display color characterization.In this study,a model of display color characterization based on Seagull Optimization Algorithm(SOA)to optimize BP neural network(SOA-BP)was proposed.Firstly,the K-fold cross-validation method was used to determine the number of the best hidden layer neurons of BP neural network,and the BP neural network model between the display input signal RGB value and the display color L^(*)a^(*)b^(*)was established.Then,the SOA was used to optimize the weight threshold of BP neural network,and the display color characterization model based on SOA optimized BP neural network was established.Finally,the neural network model before and after optimization was trained and tested,and the model accuracy was compared and analyzed.In the simulation experiments,the average values of CIELAB color difference ΔE_(ab)^(*) and CIE2000 color difference ΔE_(00) obtained by the BP neural network prediction model optimized by SOA for 10 times are 2.291 and 1.032,respectively,which are reduced by 31.79%and 36.18%compared with before optimization,and the optimized SOA-BP model is more stable.It showed that,the SOA-BP network model established in this study has higher prediction accuracy and better stability for the display color characterization,which provides a theoretical and practical reference for the color management color characterization in the printing and packaging field.

关 键 词：BP神经网络 显示器颜色特性化 K折交叉验证 海鸥算法 

分 类 号：TS801.3[轻工技术与工程] TP183[自动化与计算机技术—控制理论与控制工程]