Identification of textile fabrics
TRICLOPS spectrometer allows a fast identification of fabrics (such as wool, cotton, and polyester) in real time and with high accuracy.
TRICLOPS spectrometer can be configured in diffuse reflectance mode to quickly and simply collect reflectance spectra of many materials.
Here we show the reflectance spectra of several wool, cotton, and polyester fabrics.
The reflectance spectra (each collected in <10 seconds) of the fabrics are clearly distinct, especially at mid-IR wavelengths beyond 2500 nm.
Spectra at these wavelengths are not affected by fabric color or weave.
These measurements were used to train an identification model.
Using this model, TRICLOPS can operate as an automatic analyzer, identifying samples made of these fabrics in real-time with high accuracy.
Introduction
The classification of textile fabrics is a critical task in various industries, including fashion, manufacturing, and recycling. Accurate identification of fabric types such as wool, cotton, and polyester is essential for quality control, product labeling, and material recycling processes.
Traditional methods of fabric identification, which often rely on visual inspection and manual testing, can be time-consuming, subjective, and prone to errors.
Infra-red spectroscopy offers a robust, non-destructive solution for fabric classification, exploiting the unique spectral features characterizing the structure of different materials, providing distinct spectral fingerprints for each type of fabric. This allows for precise and rapid identification.
Measurements
An experimental analysis was conducted over a total of 9 textile samples, grouped as pure wool, pure cotton, and pure polyester. Two spectra for each sample were taken. The different samples within each fabric featured various colors and weave structures, presenting possible confounding factors that could cause misidentification.
A simple diffuse reflectance setup was used for the measurements, illuminating the samples with the focused light of a halogen lamp, and collecting the diffuse signal with a lens positioned specifically to avoid the specular reflection signal.
Experimental results
The spectra were obtained using TRICLOPS in full detection range (400–4500 nm) with fine (10 cm-1) resolution averaging 20 scans (10 seconds total collection time) per measurement. The corresponding spectra are shown in Figure 5 a).
A uniformly reflecting reference (spectralon) was used as a blank to convert the raw spectra to absorbance. The spectra were then pre-processed (normalized and baseline corrected). The pre-processed spectra are shown in Figure 5 b).
The three fabrics present very distinct spectra, especially at wavelengths >2600 nm. Furthermore, the spectra of each fabric are clearly very consistent. Using only infra-red regions, we avoid variation due to sample color or weave.
Figure 5: a) Raw spectra from TRICLOPS b) Pre-processed spectra
Data analysis
A PLSDA (Partial Least Squares Discriminant Analysis) algorithm was chosen for developing a reliable predictive model that can univocally classify different fabrics. The validation of this model was done by 5-fold cross validation (CV): 5 subsets of the dataset were created to sequentially calibrate the model on 4 of them and evaluate its predictive effectiveness on the remaining one.
An additional analysis of the single wavelengths importance in the regression model was done to reduce the predictors’ set, improving the model performances.
The figures of merit for the model’s evaluation are the accuracy of cross validation and the correlation coefficient R^2. The results are shown in the table below.
To visually observe the clustering effectiveness of the PLS algorithm on the different fabrics the PLS scores plot is shown in Figure 3.
Figure 6: PLS scores plot
Conclusion
In this application note the effectiveness of TRICLOPS for the classification of textile fabrics. With a short and simple set of measurements, a model was developed which correctly identified the fabrics with high accuracy. This is helped by the long wavelength sensitivity of TRICLOPS, which gave access to particularly rich signals in the region from 2700–4000 nm.
Discover TRICLOPS Spectrometer
