Effect of humidity on fabric color difference based on a hyperspectral imaging system and convolutional neural network

Abstract

The effect of humidity on fabric color characteristics is an important research area, because humidity has a major effect on fabric color characteristics. In this paper, the effect of different humidity levels on fabric color characteristics was studied using a grating hyperspectral imaging system. This work provides a detailed analysis of the changes in surface spectral reflectance of fabrics under different humidity levels. The reflectance of fabrics decreases with increasing humidity and exhibits similar reflective characteristics. This work also calculated the color difference value of the fabric and found that the color difference value of the fabric increases with the increase of humidity. At the same time, a one-dimensional (1D) convolutional neural network (CNN) prediction model based on hyperspectral data was established, with the spectral reflectance of fabrics under different wavebands as input and the fabric color difference value as output. In addition, in order to improve the prediction accuracy of the model, this study used Bayesian parameter optimization algorithm to optimize the model parameters and obtain the optimal combination of model parameters. At the same time, decision tree, random forest, and support vector machine models were built for comparison, and the experimental results show that 1D-CNN has the lowest average prediction deviation of 0.2728 among all samples, which demonstrates the superiority of the model prediction performance.