Color prediction model for 3D printing based on multi-layer BP neural network

Color reproduction of 3D printing is affected by several factors including materials, morphology, viewing conditions, etc. In this paper, color prediction model based on multi-layer BP neural network was proposed for color reproduction of photosensitive resin for 3D printing. In such model, three main factors, the digital color values, the background factors for measurement and the thickness of protected layer for measurement, were chosen to work as the main control factors. Ceramic camera calibration plate and 50% neutral grey blocks were selected as the measurement background, which facilitated to control and analyse the impacts of the background on the measurement. And clear translucent blocks of different thickness, 2mm, 4mm, 6mm and 8mm, were deposited on the 3D printed color blocks surfaces, which facilitated to analyse the impact of light scattering on the measurement and color appearance. Three evaluation methods, the mean absolute error (MAE), the coefficient of determination(R²), and the average color difference(ΔE*ab), were used to evaluate the performance of prediction model. It shows that the prediction model based on multi-layer BP proposed in this paper has a higher prediction accuracy. Furthermore, the protected layer thickness and background factor both have obvious effects on color reproduction of 3D printing according to the weight of each factor