The residual moisture content of black textiles finished with functional agents by impregnation was monitored after drying by near-infrared (NIR) hyperspectral imaging. Quantitative moisture data are inevitably required for optimum control of the drying process of wide textile webs during converting. Since spectral imaging is carried out in reflection, the very low signal intensities resulting from both the low reflectance of black materials and the usually rough surface structure of textiles leading to scattering of the incident probe light make exceedingly high demands on both the measurement instrumentation and the multivariate regression methods used for quantification of the spectral data. The black color was found to have a strong effect on the reflectance spectra, which became manifest in a marked background. Nevertheless, the best prediction performance was obtained for calibration models based on spectra without removal of the background. External validation of the models with independent samples yielded a root mean square error of prediction (RMSEP) of 0.55 wt%. Surprisingly, this prediction error is completely in line with the corresponding errors obtained previously for comparable bright textiles in spite of the much more challenging optical conditions for the black materials. Using the developed calibration models, hyperspectral monitoring of finished black blended fabrics with both homogeneous and nonhomogeneous moisture distribution was carried out. It was shown that both local and large-range fluctuations can be easily detected by spectral imaging, which makes this method very useful for the control of technical textile production and converting processes.
Read full abstract