一种基于RBF神经网络的打印机光谱预测模型

本文提出一种基于RBF(Radial Basis Function,径向基函数)神经网络的打印机光谱预测模型,通过扩展神经网络模型输入变量的项数提高模型的预测精度,扩展项多采用通道驱动值的交叉值、平方值.实验结果表明[1cmy]项的引入能够有效提高模型的预测精度,同时提高网络的泛化能力.而引入[cm² cy² mc² my² yc² ym²]项会导致模型预测精度以及泛化能力降低.[1 cmy]、[c²m²y²]和[cm cy my]项的组合在预测精度和模型泛化能力上均是最优化的,对总样本预测的色度精度为0.475ΔE₀₀,光谱精度RMSE为0.43%.因此选择[1 cmy c²m²y² cm cy my c m y]作为输入变量的RBF神经网络训练模型是满足高精度光谱预测的最优模型.

A Spectral Prediction Model of Printer Based on RBF Neural Network

A spectral prediction model of printer based on RBF (Radial Basis Function) neural network was proposed in this paper. Prediction accuracy of model is improved by extending the input variables of neural network. The extensions included cross value or square value of channel driven value. Experimental results show that introduction of [1 cmy] item can effectively improve the prediction precision and the generalization ability, introduction of [cm² cy² mc² my² yc² ym²] item can decrease the prediction precision and generalization ability of the network. Introduction of combination of [cmy]、[c²m²y²] and [cm cy my] item can achieve the optimized prediction accuracy and generalization ability. This combination terms can reach the colorimetric accuracy of 0.475 ΔE₀₀ and spectral accuracy of 0.43% RMSE of all samples. Thus, RBF neural network model with input variables of [1 cmy c²m²y² cm cy my c m y] is the most optimized model that meets spectral prediction with high resolution.