Custom hyperspectral imaging scanner for microplastic detection and classification: hardware and data processing specifications

This paper describes a micro-spectral scanner designed for microplastic recognition and an optimized data processing approach for their identification and classification. Microplastics represent an increasing threat to marine and terrestrial ecosystems, accumulating in the food chain and posing potential risks to human and environmental health. Advanced and precise techniques are needed to identify and classify these particles at a microscopic level. However, current analytical methods require long acquisition and processing times, limiting large-scale analysis and operational efficiency. Hyperspectral imaging (HSI) offers an effective solution by leveraging the distinct spectral signatures of microplastics for detailed, non-destructive analysis on a large scale. Nonetheless, the observation of microscopic objects like microplastics requires balancing high resolution, necessary for identifying fine details, and the ability to analyze large sample quantities efficiently. When prototyping HSI devices for microplastic investigation, challenges related to optics, imaging acquisition and instrumentation must be addressed. In pushbroom systems, where scanning occurs line by line, spectral mapping accuracy presents a significant challenge. Since the image is acquired gradually, any variation in acquisition speed, microvibrations, or movement can cause distortions or discrepancies in the data. Additionally, the type and geometry of illumination significantly affect reflectance and the signal received by the sensor, influencing the signal-to-noise ratio. Finally, a specific chemometric approach is essential to optimize the analysis of acquired data. In this work the developed micro-HSI scanner is described and its performances are evaluated through specific tests carried out on selected microplastics of different polymers and sizes, discussing its challenges and limitations.