基于BP神经网络的彩色3D打印颜色预测模型

Color Prediction Model for 3D Printing Based on BP Neural Network

彩色3D打印颜色再现受到透明、半透明色块颜色等多方面因素的影响。本研究基于BP神经网络,建立了透明树脂厚度、底基、标准24色块的输入值与3D打印色块颜色值之间的颜色预测模型。将4mm厚的3D打印24色色块作为测量对象,陶瓷相机标定板和50%中性灰的铜版纸作为测量底基,2mm、4mm、6mm、8mm厚度的无色透明树脂块分别置于彩色3D打印色块表面,用分光光度计测量色块的色度值和明度值,并用Matlab R2020a建立3D打印输入值与测量值的关系模型,以实际测量值与预测值的色差评价预测效果,以输入参数权重的大小判断3D打印控制参数对最终结果影响的强弱。实验结果表明:验证集的总色差最大值为4.855,平均值为1.974。透明树脂厚度的权重为0.0491、测量底基的权重为0.0869,依据权重信息可知,测量底基对预测结果影响最大,无色透明树脂块的厚度影响次之。本研究建立的颜色预测模型具有较好的预测效果。

With the development of color 3D printing technology,color reproduction has attracted more and more attention,and its color reproduction is affected by many factors,such as the transparent or translucent color of the 3D printed color block.In this study,based on BP neural network,a color prediction model was established between the thickness of transparent resin,the subbase,the standard 24 color block and the 3D printed color block.First,3D printed 24 color block of 4mm thickness as the measurement object,ceramic camera calibration plate and 50%neutral gray copper paper as the measurement base,the transparent resin blocks of 2mm,4mm,6mm and 8mm thickness were placed on the surface of the color block to measure the color value and lightness value of the color block with the spectrophotometer.The relationship model between the 3D printing input value and the measured value were established by Matlab R2020a,the prediction effect was evaluated by the color difference between the actual measured value and the predicted value,and the influence of the 3D printing control parameters on the final result was judged by the size of the input parameter weight.The results showed that the maximum total color error was 4.855,and the mean color error was 1.974.V of thickness is 0.0491,and V of subbase is 0.0869.According to V,the subbase has the more influence on the prediction results,and the thickness of colorless transparent block has the less influence on the prediction results.The color prediction model established shows a good prediction effect.