基于 BP 神经网络的多基色打印机光谱特性化

目的实现多基色打印机的光谱特性化。方法结合光谱降维和光谱重构方法建立了多基色打印机光谱特性化BP神经网络模型,并提出了基于人眼视觉特性加权的目标函数。结果在基于BP神经网络的多基色打印机光谱特性化中,当目标函数未进行人眼视觉特性加权时,光谱精度和色度精度分别为0.0285和2.8614,当采用人眼视觉特性加权目标函数后,光谱精度和色度精度分别为0.0166和1.2247。结论在基于BP神经网络的多基色打印机光谱特性化中,使用基于人眼视觉特性加权的目标函数可兼顾光谱与色度2个因素,其光谱特性化效果更优。

Spectral Characterization of Multicolor Printer Based on BP Neural Network

Objective To achieve the spectral characterization of multicolor printer. Methods Combining spectral dimensionality reduction and spectral reconstruction method, we built a BP neural network model for multi-color printer spectral characterization, and proposed an objective function based on human visual characteristics.Results In the multi-color printer spectral characterization based on BP neural network model, when the objective function was without human eye visual feature weighting, the spectral accuracy and chromaticity accuracy were 0.0284 and 2.8614, while the objective function employed human eye visual feature weighting, the spectral accuracy and chromaticity accuracy were 0.0166 and 1.2247, respectively. Conclusion In the spectral characterization of multicolor printer based on BP neural network, using objective function based on human eye visual feature weighting could cover both factors of spectrum and chromaticity, and its spectral characterization effect was better.