Automated Quantification of Microplastics – Challenges and Opportunities

Plastics are an important material with widespread applications. However, their widespread use and poor end-of-life management have led to their extensive environmental pollution. They can be found in oceans, terrestrial ecosystems, and even remote corners of the Earth. Current methods for microplastic quantification and identification require big investments and highly trained personnel to operate the analytical equipment. In this paper, we propose an algorithm-based method for the quantification of microplastics in soil and organic fertilisers. The method is based on image analysis of a thinly spread sample that was heated until microplastics has visually melted. The algorithm-based method was validated with Focal plane array detector-based micro-Fourier-transform infrared imaging (FPA-μFTIR), frequently used in microplastic characterisation. Herein, we present the pre-liminary results of an ongoing study. In a compost sample, five particles were detected with FPA-μFTIR, whereas the algorithm detected eight. The algorithm has difficulties recognising elongated or oddly shaped particles. These were identified as several particles which led to overestimating the number of microplastic particles in the investigated sample. We will continue with further develop-ment of the computer algorithm by using a training set of images which will be quantified using different methods (visual detection by a human operator, FPA-μFTIR). This growing training set will enable us to incorporate machine learning algorithms (neural networks) in the development of a more reliable particle detection algorithm. We expect that environmental monitoring of microplas-tics will be required under future legislation, therefore the development of cheap, user-friendly so-lutions is crucial. Keywords: Machine learning; Algorithm; Infrared spectroscopy; Soil contamination; Organic ferti-lisers; Compost