SE-CNN Intelligent Color Matching Model Based on Kubelka-Munk Single Constant Theory

Now, the majority of investigations use traditional neural networks to realize intelligent color matching, which is prone to over-fitting and the generalization ability is weak. We explore the feasibility and effectiveness of the method of that attention mechanism combined with a convolution neural network to solve the problem of intelligent color matching based on the Kubelka-Munk optical model (K-M optical model), color mixing law, and Kubelka-Munk single constant theory (K-M single constant theory). Through the above theory, we got the relationship between the reflectance (R) of the theoretical substrate and the concentration of pigments. Theoretical reflectance was generated from the pigment concentration. reflectance and concentration were used as the data set in this paper. Through using the attention mechanism, Squeeze-and-Excitation Block (SE block), build SE-CNN intelligent color matching model, realizing the predictions of the dyeing formula. Each resulting average relative error of the predicted recipe and real recipe is 2.81%,2.71%,2.81%, and the relative errors are all less than 10.5%.80% of them are less than 5%. The network model can be used to solve intelligent color-matching problems and provide a new direction improving the effect of intelligent color-matching algorithm.