基于改进PSO-Elman的液晶显示器颜色特性化  

Color Characterization of LCDs Based on Improved PSO-Elman

作　　者：孙士明[1] 倪潇 李媛媛 高绍姝[1] SUN Shi-ming;NI Xiao;LI Yuan-yuan;GAO Shao-shu(Qingdao Software College,China University of Petroleum(East China),Qingdao Shandong 266580,China;Daqing Power Supply Bureau,Heilongjiang Power Grid Co.,Ltd.,Daqing Heilongjiang 163453,China)

机构地区：[1]中国石油大学(华东)青岛软件学院,山东青岛266580 [2]国网黑龙江省电力有限公司大庆供电公司,黑龙江大庆163453

出　　处：《计算机仿真》2024年第6期274-279,286,共7页Computer Simulation

基　　金：国家自然科学基金项目(61801517);中央高校基本科研业务专项经费(19CX02029A,19CX02027A)。

摘　　要：液晶显示器颜色特性化可以实现同一幅图像在不同设备上的准确显示。为解决液晶显示器颜色特性化存在模型建立复杂、模型鲁棒性差导致特性化精度较低的问题,提出基于改进PSO-Elman神经网络的方法建立RGB颜色空间到CIEXYZ颜色空间的转换模型(ACOPSO-Elman)。首先根据粒子种群规模和粒子位置关系构造惯性权重与学习因子的自适应调节函数提高PSO算法的全局寻优能力和收敛速度,并在寻优过程中添加混沌优化(CO),防止粒子陷入局部最优解,将改进的粒子群算法用于Elman模型参数寻优,解决了Elman模型参数较难选取的问题。通过仿真验证并与BP、Elman神经网络模型比较表明,ACOPSO-Elman模型特性化的平均色差为1.9247ΔE^(*)_(ab),最大色差为5.1252ΔE^(*)_(ab),在特性化精度上取得了较好的效果。The LCD color characterization can realize the accurate display of the same image on different devices.To solve the problems of complicated model establishment and poor model robustness leading to low characterization accuracy in LCD color characterization,a conversion model from RGB color space to CIEXYZ color space(ACOPSOElman)is proposed based on an improved PSO-Elman neural network approach.Firstly,the adaptive adjustment function of inertia weight and learning factor is constructed according to the particle population size and particle posi⁃tion relationship to improve the global optimization-seeking ability and convergence speed of PSO algorithm,and chaos optimization(CO)is added in the optimization-seeking process to prevent the particles from falling into local optimal solutions,and the improved particle swarm algorithm is used for Elman model parameter-seeking to solve the problem of difficult selection of Elman model parameters.The average color difference of the ACOPSO-Elman model characterization is 1.9247ΔE^(*)_(ab) and the maximum color difference is 5.1252ΔE^(*)_(ab),which achieves better results in the characterization accuracy,as verified by simulation experiments and compared with BP and Elman neural network models.

关 键 词：神经网络 液晶显示器 颜色特性化 粒子群算法 自适应调节函数 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]