Light consistency correction for the liquid crystal tunable filter hyperspectral imaging system

Abstract

In hyperspectral images, every pixel encompasses continuous spectral information. Compared with traditional colorimeters, using hyperspectral imaging systems (HIS) for fabric color measurement can result in obtaining richer color information. However, measuring fabric colors with liquid crystal tunable filter HIS can lead to challenges related to light consistency. In this paper, we adopted an innovative approach, integrating gradient boosted decision trees with a sliding window algorithm to develop a uniformity calibration model addressing the illumination uniformity issue. To address the consistency issues across various light sources, we further adopted a deep neural network (DNN) model to correct the reflectance measurements under different light sources. Subsequently, this model was merged with the uniformity calibration model to form a light-consistency correction model. Through calibration, we successfully reduced the color difference of the corrected samples from 3.636 to 0.854, an enhancement of 76.51%. This means that after calibration we can achieve consistency in fabric color measurements under nonuniform lighting and different light sources.