Conclusions | Source |
---|---|
Water reduces spectral reflectance, drying of particles is important | |
SVM model outperforms PLS-DA, ML | |
Dark particles are harder to detect | |
ID of smaller particles is hindered by spatial resolution | |
Weathered polymers have degraded spectra as compared with pristine polymers | |
Reference polymers should be more diverse, include weathering and polymer mixtures | [65] |
Over 90% of particles > 1 mm positively identified | |
Addition of extra “class” for background or non-polymer material helped classification | |
Glass fiber filter is a suitable background material |