Hyperspectral Imaging for Microplastic Detection Using Specim FX17 and SWIR Cameras

Microplastics have become pervasive in both aquatic and terrestrial environments. These particles, typically ranging from 10 to 500 micrometers, originate from the degradation of larger plastic materials. Their small size allows them to spread widely and be unknowingly ingested by organisms, including humans, raising serious concerns about their potential impact on health. Reliable detection and identification of microplastics are essential to better understand their formation pathways and to quantify their presence in natural ecosystems.

To support this effort, the Norwegian Institute for Water Research (NIVA) supplied a range of plastic samples for spectral analysis using Specim’s hyperspectral imaging technology. The samples consisted of larger granules measuring a few millimeters each as well as filtered microplastics of several sizes. The goal was to assess the capability of Specim’s FX17 and SWIR cameras in identifying microplastics and to evaluate whether spectral data collected from larger plastic granules could be applied effectively to detect smaller microplastic fragments.

Sample Description

The study included a diverse set of plastic materials (Figure 1). Larger granulates, each a few millimeters in size, were provided as base samples to build a spectral reference library. These consisted of commonly used polymers such as high-density and low-density polyethylene (HDPE and LDPE), polyethylene terephthalate (PET), polycarbonate (PC), polypropylene (PP), polyamide (PA), two types of polystyrene (PS1 and PS2), and polyvinyl chloride (PVC). These materials are frequently found in the environment due to their widespread use and tendency to degrade into microplastics over time.

In addition to the macro samples, microplastic particles made from polyethylene (PE) and polystyrene (PS) were also included in the analysis. These microplastics varied in size and color and served to evaluate the performance of the spectral library when applied to smaller particles.

Figure 1: Photos of samples.

Measurement Setup

Spectral reflectance measurements were carried out using Specim FX17 and SWIR hyperspectral cameras, covering spectral ranges of 900–1700 nm and 1000–2500 nm, respectively. For the larger plastic granules, the Specim FX17 was equipped with a 38-degree lens, and the SWIR camera used a 17-degree lens, resulting in a field of view of approximately 10 cm. The corresponding pixel sizes were approximately 0.2 mm for Specim FX17 and 0.3 mm for SWIR.

For the microplastic measurements, a high-resolution imaging setup was implemented using the Specim OLES Macro lens. This setup allowed for pixel sizes of 19 micrometers with Specim FX17 and 24 micrometers with the SWIR camera, enabling the detection of fine details in very small particles. All measurements were conducted using the LabScanner 40 x 20 platform, and data processing was carried out with SpecimINSIGHT software.

Spectral Analysis: The Role of Sample Size

A key objective of the study was to determine whether a spectral library built from larger plastic particles could be reliably used to identify microplastics. The analysis focused specifically on polyethylene (PE) and polystyrene (PS), the two polymers present in both the macro and micro samples. Hence, the samples and spectra cover only these materials.

Figure 2: Specim FX17 reference spectra for PE and PS, also related to a single microplastic per size.

The results revealed that as the size of the sample decreases, its spectral features become less distinct (Figures 2 and 3). This effect is particularly noticeable in polystyrene microplastics, which were generally smaller and more transparent than the polyethylene samples. As a result, the spectral signatures of these materials become less pronounced, making their identification increasingly difficult at smaller scales.

The SWIR camera covers an extended spectral range of 1000–2500 nm, allowing it to detect features that are not visible with the Specim FX17, which operates in the 900–1700 nm range. This provides more comprehensive spectral information that aids in the identification of these challenging samples.

Figure 3: SWIR reference spectra for PE and PS, also related to a single microplastic per size.

Modeling and Identification of Microplastics using Hyperspectral Imaging

To evaluate detection performance, a classification model was developed using partial least squares discriminant analysis (PLS-DA). The reference spectral library built from the larger samples was applied on the microplastics by using this model. For simplicity and improved classification, HDPE and LDPE spectra were grouped into a single PE class, and both PS1 and PS2 were combined into a single PS class.

Specim FX17 camera

When applying the model with the Specim FX17 camera, microplastic particles were generally well identified (Figure 4a). The spectral data from the larger samples transferred effectively to the smaller particles, demonstrating the potential of this method.

However, some misclassifications were observed. Certain polyethylene (PE) particles were incorrectly labeled as polypropylene, and some polystyrene (PS) particles were identified as PVC. These errors were more pronounced among the smallest microplastic particles, where reduced spectral features limited the model’s discriminative abilities. In addition, several PE particles went undetected altogether, suggesting a limitation in sensitivity for extremely small or translucent particles (Figure 4b).

Figure 4a: Modeling results obtained with Specim FX17.

Figure 4b: Modeling results obtained with Specim FX17, zoomed on the smallest particles.

SWIR camera

In contrast, the SWIR camera demonstrated improved classification accuracy. Despite having slightly larger pixels compared to the Specim FX17 (24 vs. 19 um), the broader spectral coverage of the SWIR camera (1000–2500 nm vs. 900–1700 nm) enabled better sorting accuracy of material types (Figures 5a and 5b).

Figure 5a: Modelling results obtained with SWIR.

Figure 5b: Modeling results obtained with SWIR, zoomed on the smallest particles.

The additional spectral information captured between 1700 and 2500 nm with the SWIR enhanced the model’s ability to detect and accurately classify both polyethylene and polystyrene microplastics. The results showed that the Specim’s SWIR hyperspectral camera provided more consistent and accurate identification across a wider range of particle sizes.

---

---

Conclusion

This study demonstrates the effectiveness of hyperspectral imaging in detecting and classifying microplastics. A spectral library developed from larger plastic granules can be applied with reasonable success to identify smaller microplastic particles, particularly when using cameras with broader spectral coverage. While the Specim FX17 camera performs well within its spectral range, the Specim SWIR system offers improved accuracy, especially for particles at the lower end of the microplastic size spectrum. These findings highlight the potential of hyperspectral imaging as a powerful tool for environmental monitoring and microplastic research.

---

---

Disclaimer

This technical note was prepared by Specim, Spectral Imaging Ltd. and for generic guidance only. Specim keeps all the rights to modify the content.