Plastics are all around us, but what really happens to them after we throw them in recycling bins? Large quantities end their life in a landfill or incinerators, never being recycled at all.
Only a small portion of plastics have value to the recycling industry, but it is costly and time-consuming to separate these high-value materials.
Labeling plastics resins for macro identification (see Fig.1 below), but this is not useful when breaking the materials into parts or when the labeling is inaccurate. In these cases, sorting is incorrect or impossible by hand, color visualization, or advanced AI and machine learning techniques.
Our environment would benefit significantly from a process that could safely, quickly, and accurately separate materials to increase recycling. In addition to saving the resources needed to produce new plastics, better sorting processes can also reduce dangerous chemical pollution in the atmosphere.
Hyperspectral imaging offers a highly effective and versatile solution for efficient plastic sorting in recycling facilities. By accurately identifying and classifying plastics based on their spectral properties, hyperspectral imaging contributes to higher-quality recycling outcomes, reduced contamination, and a more sustainable approach to plastic waste management.

Figure 1: Plastics labeling and code number.
Hyperspectral imaging relies on plastic resin’s chemical composition and the recycling industry has increasingly adopted it recently. The capability of hyperspectral cameras to sort plastics is not new. However, some limitations have prevented the widespread implementation of this technology in plastic recycling, up until now:
- Hyperspectral cameras were slow, which meant a low return on investment.
- Spectroscopic methods could not sort black plastics.
- Cameras were prohibitively expensive.
With the Specim FX series hyperspectral cameras, we’ve broken these hurdles and opened the door for new industrial hyperspectral imaging applications. The Specim FX cameras are fast, affordable, and cover an extensive spectral range. Specim is the only hyperspectral camera manufacturer dedicated to covering the full spectral range of 400 – 5200 nm needed for the industry. Specim FX10 covers the spectral range of 400 – 1000, Specim FX17 covers 900 – 1700, the SWIR covers 1000 – 2500, and Specim FX50 covers 2700 – 5200 nanometers.
Specim is the leading global provider of hyperspectral imaging solutions, especially for the recycling industry, with many Specim FX cameras used in recycling plants for sorting different types of plastics.
We have studied many plastic sample types extensively across the complete spectral range. We can confidently recommend the best HSI camera solution based on the plastic type and application of the end user.
Our studies included measurement, analysis, and classification of PE, ABS, PVC, PS, PA, PP, PC, and PET. Those are the most commonly recycled plastics (and labeled as in Fig.1). The samples were mainly white or transparent and measured in the Specim laboratory.

Figure 2: Example of plastics samples measured.
We processed the data with our SpecimINSIGHT software which is a part of the SpecimONE Spectral Imaging Platform*, and built PLS-DA models for each data cube. We limited the spectral range to 700 – 1000 nm of the Specim FX10 to remove any bias in the sample’s color.
Figure 3. Sorting results for each camera; Green: PE; Red: ABS; Blue PVC; Yellow: PS; White: PA; Sea green: PP; Cyan: PC and Purple; PET.
As can be seen in Fig.3, in general, sorting was relatively efficient.
- For Specim FX10, we can sort most of the samples. Notice that PC is usually transparent, so only the edge can be sorted with the center misclassified as background.
- For Specim FX17, we produced an accurate and robust model. Same remark as for FX10 regarding the samples made of PC
- For the SWIR camera, we produced the most accurate model. All samples were well sorted, even the one made of PC.
- The model based on the Specim FX50 data was not as robust and accurate as the ones built with the SWIR and FX17 cameras but is still very relevant. We also could sort transparent PC.
The spectral signature of all these samples can be found in Fig.4.

Figure 4: full spectra (400 – 5300 nm) measured by Specim cameras of the most common plastics, i.e. PE, ABS, PVC, PS, PA, PP, PC and PET.
It is well known that black plastics are problematic and cannot be sorted with traditional near-infrared (NIR) cameras. However, black plastics are widely used in the automotive and electronics industries, so sorting and recycling are vital.
The Specim FX50 mid-wave-infrared (MWIR) hyperspectral camera is the only solution available for sorting black plastics. To demonstrate the power of Specim FX50, we measured similar samples as those presented previously, but with carbon contained within the polymers to make them black. Again, we used the SpecimINSIGHT software to build PLS-DA models for data classification. Results are shown in Fig.5 a with PS, ABS, PE, PA, and PC samples: We sorted all plastic types.

Figure 5.
In summary, Specim FX cameras are suitable for sorting plastics, regardless of their color. Depending on the application requirements, different cameras can be used. Table 1 below summarizes the results.

Table 1: plastics and the sorting ability of each camera.
* PE, LDPE and HDPE can not be separated.
** Black plastics can also be sorted
* PE, LDPE and HDPE can not be separated.
** Black plastics can also be sorted
*SpecimONE spectral imaging platform with data analysis software and real-time data processing makes developing new hyperspectral imaging applications easy.