Back to papers
0
Article ? AI-assigned paper type based on the abstract. Classification may not be perfect — flag errors using the feedback button. Tier 2 ? Original research — experimental, observational, or case-control study. Direct primary evidence. Detection Methods Sign in to save

Random Forest single channel classifier for FTIR Microplastic images

Zenodo (CERN European Organization for Nuclear Research) 2022
Jordi Valls Conesa, Dominik J. Winterauer, Niels Kröger‐Lui, Sascha Roth, Stephan Lüttjohann, Roland Harig, Jes Vollertsen

Summary

Researchers applied random decision forest machine learning to identify microplastics from Fourier-transform infrared (FTIR) spectral data, using a small number of highly informative wavelengths to classify particles accurately and rapidly. Faster automated spectral classification could significantly increase the throughput of microplastic analysis in environmental monitoring.

In this poster we intend to present our latest results on fast Microplastic (MP) detection. Random decision forests (RDFs) are used to build multiclass models for fast identification of Fourier-transform infrared (FT-IR) spectra of MP most common in environmental samples. The RDF operates with a reduced number of highly discriminative wavenumbers, their selection has been optimized through machine learning to select the most significative wavenumbers. Single wavenumber allows for input from direct IR, decreases inference time, and increases classification accuracy. The training and validation data are extracted from FT-IR data of purpose-made pure-type MP samples using reference spectra and the fast background correction and identification (FBCI) algorithm. RDF classification results are validated with procedurally generated MP samples as ground truths Also see: https://micro2022.sciencesconf.org/425950/document

Read via DOI Download PDF

Share this paper

Post Share Share Pin Email