Oil is used in many industries as a lubricant. In the final product, oil is usually considered a pollutant, which is crucial to detect. However, oil is complicated to see with the human eye, and therefore traditional RGB cameras are not capable of detecting it. Hyperspectral cameras, on the other hand, should be able to capture this information when operating on the right wavelength.

To test this, we spread three different types of oil on pieces of aluminum and black fabric (see Fig 1.) and scanned them with three different kinds of Specim hyperspectral cameras: FX17, SWIR, and FX50. For the test, we used Weldlite TF2, a very common lubricant used, e.g., for bike chains, Würth HSP 1400, a high-temperature lubricant, and Pentisol, synthetic oil for general use.

  • Specim FX17 (900 – 1700 nm)
  • Specim SWIR (1000 – 2500 nm)
  • Specim FX50 (2700 – 5300 nm)
Figure 1: The three types of oil used in this study and spread on the metal and fabric pieces. The circles on the metal plate and piece of fabric mark the area where the oil was applied. Green indicates the Weldlite, red the Würth, and blue the Pentisol oil.

Specim FX17 camera

The Specim FX17 camera covers the 900 – 1700 nm spectral range and is widely used in industrial quality control. It is suitable for detecting various chemical substances based on their natural and synthetic compounds. For example, it is used to measure the quantity of substance, e.g., nicotine in tobacco leaves, and to detect contaminants and unwanted objects, e.g., pieces of bones in the minced meat. Concerning mechanical oils, very little has been tested so far.

Based on the spectral analysis (see Fig 3. and 4.), Specim FX17 camera can not detect all oils.
The Pentisol oil can be somewhat detected, the Würth oil can be detected mostly on fabrics (absorption peak at 1393 nm.), and the Weltlite oil cannot be detected at all.

Figure 2: A false RGB image of the samples with selections.
Figure 3. Spectra measured with Specim FX17 camera related to the selections averaged. Green indicates the Weldlite, red the Würth, and blue the Pentisol oil. The yellow spectra can be considered the reference as they are only related to the metal without being spoiled by any type of oil.
Figure 4. Spectra measured with Specim FX17 camera related to the selections averaged. Green indicates the Weldlite, pink the Würth, and blue the Pentisol oil. The orange spectra can be considered the reference as they are only related to the fabric without being spoiled by any type of oil.

A Principal Component Analysis (PCA) was carried out on the data, confirming the observations based on spectra. The analysis further shows that the oils that Specim FX17 can detect can be sorted from each other very faintly (Fig 5.).

Figure 5: A false representation of the PCA by assigning PC1 to the red, PC2 to green, and PC3 to the blue band.

Specim SWIR camera

The Specim SWIR camera covers the 1000 – 2500 nm spectral range. Like the FX17, SWIR is also suitable for detecting different chemical components. Due to its wider spectral range, SWIR detects more materials than the FX17.

Based on the spectral analysis (Fig 7. and 8.), the SWIR camera, with spectral features above 2200 nm, is suitable for detecting the different types of oil, especially the Würth and Pentisol oils. The Weltlite oil can be partly detected.

Figure 6: False RGB image of the samples with selections.
Figure 7. Spectra measured with Specim SWIR camera related to the selections averaged. Green indicates the Weldlite, red the Würth, and blue the Pentisol oil. The yellow spectra can be considered the reference as they are only related to the metal without being spoiled by any type of oil.
Figure 8. Spectra measured with Specim SWIR camera related to the selections averaged. Green indicates the Weldlite, pink the Würth, and blue the Pentisol oil. The orange spectra can be considered the reference as they are only related to the fabric without being spoiled by any type of oil.

In combination with a PCA, a Partial Least Square–Discriminant Analysis (PLS-DA) was carried out to evaluate the sorting performance of the SWIR camera (see Fig 9.). Based on the results, Würth and Pentisol oils can be detected from both metal and fabric surfaces and sorted from each other with the Specim SWIR camera. The Weltlite oil can be detected on fabrics, but it cannot be detected in a reliable manner on metal.

It is noteworthy that, even though all the oils were applied to very restricted areas of the samples, the Weltlite oil has spread extensively on the fabric.

Figure 9: Left: false representation of the PCA by assigning PC1 to the red band, PC2 to green, and PC3 to blue; Right: PLS-DA model prediction.

Specim FX50 camera

The Specim FX50 camera covers the 2700 – 5300 nm spectral range. These wavelengths are part of the thermal range, so-called MWIR. The FX50 is very suitable for sorting different types of polymers, regardless of their color – even black ones.

The results show that the FX50 can detect all three different types of oils both on the metal and the fabric surface. FX50 can detect even tiny droplets of oil. The test shows powerful oil absorption between the 3300 and 3500 nm spectral range (Fig 11. and 12.).

Figure 10: false RGB image of the samples with selections.
Figure 11. Spectra measured with Specim FX50 camera related to the selections averaged. Green indicates the Weldlite, red the Würth, and blue the Pentisol oil. The yellow spectra can be considered the reference as they are only related to the metal without being spoiled by any type of oil.
Figure 12. Spectra measured with Specim FX50 camera related to the selections averaged. Green indicates the Weldlite, red the Würth, and blue the Pentisol oil. The yellow spectra can be considered the reference as they are only related to the fabric without being spoiled by any type of oil.

In addition to the PCA, a PLS-DA modeling was built to assess the sorting performance of the Specim FX50 camera. As can be seen in figure 13., all three types of oils can be detected from both metal and fabric surfaces and sorted from each other with the Specim FX50 camera.

Figure 13: Left: false representation of the PCA by assigning PC1 to the red band, PC2 to green, and PC3 to blue; Right: PLS-DA model prediction.

Conclusion
Based on the analysis, we can conclude that the Specim FX50 is the best camera to detect oils from their surface. With FX50, you can also sort different oil types from each other.

Disclaimer
Specim, Spectral Imaging Ltd., made this technical note for general guidance only. We keep all the rights to modify the content.

Glossary
PC = principal component
PCA = principal component analysis
SWIR = short waves infrared
MWIR = Mid waves infrared
PLS-DA = partial least square – discriminant analysis