Example 1: Textilies

In this example it should be shown that different textiles can be distinguished on the basis of the material. For this purpose measurements are made with polyester, cotton, viscose and nylon. The aim is that the material of any textile can be predicted.

Adobe Acrobat Dokument 428.5 KB

1. Raw data

  • Calibration measurements must be performed in a first step.
  • Each textile is measured in three different places with the NIRScreen.
  • This results in 12 calibration measurements in total.
  • When looking at the measurement results, it is noticeable that the data differed both within a material group and between the materials.
  • Thus, it is not yet clear on basis of the raw data whether the textiles can really be distinguished.

Raw data:

Cotton (red), Polyester (blue), Viscose (green), Nylon (orange)

2. Processed data

  • In the second step, the raw data are processed with an algorithm, e.g. to correct scatter effects.
  • The processed data show that the measurements within a material group are now very close together.
  • The measurements between material groups are far apart.
  • It can therefore be assumed that the textiles can be distinguished.

Processed data:

Cotton (red), Polyester (blue), Viscose (green), Nylon (orange)

3. Principal Component Analysis

  • As a third step, a principal component analysis is performed, which should also provide information on whether the textiles can be distinguished.
  • The result of the principal component analysis shows that the measurements can be divided into four different groups.
  • This confirms the impression of the processed data that the textiles can be distinguished.

Result of Principal Component Analysis:

Cotton (red), Polyester (blue), Viscose (green), Nylon (orange)

4. Statistical model

  • As a final step, a statistical model is created, with which it is possible to predict the material of a textile.
  • These statistics models are stored in the cloud and can be downloaded from a NIRScreen.
  • From the statistics model it can be read on the basis of model parameters that the textiles can be distinguished very well.
  • The statistical model also shows how strong the influence of wavelengths is on the differentiation of textiles (see graph).

Influence of wavelengths on the differentiation of textiles