Plastics have many unrivaled properties and are thus indispensable materials with an ever-increasing production. Since the plastics recycling rates are still small, the plastic waste that ends up on landfills, in the environment including oceans, and incineration plants also steadily grows. Large-scale automated sorting of waste plastics may help solve the problem. However, most state-of-the-art approaches are limited to a (very) small number of plastic types that can be simultaneously sorted and have severe difficulties with black plastics. The labeling of plastics with fluorescent tracers ("markers") during their production has the potential to enable the fast and highly reliable classification and automated sorting of many different plastic types simultaneously. Unfortunately, the unique marker fluorescence ("optical fingerprint") might be obscured by the autofluorescence (AF) often emitted by plastics. Recently, we have shown that time-gated fluorescence spectroscopy (TGFS) can be successfully applied to eliminate the influences from AF. In this study, we present the prototype of a TGFS measurement system for the classification and sorting of plastics and evaluate its performance in a typical industrial situation. With 150 000 investigated plastics flakes from 10 different plastic types that were labeled with binary combinations of six fluorescent markers, an overall plastics recovery rate (mean true positive rate ) of 99.76%, and a plastics purity (positive predictive value ) of 99.88% were achieved. Additional simulations that were carried out based upon the experiments allow a prediction of the performance as a function of the signal-to-noise (S/N) ratio such that various system parameters can be adapted to the particular situation.
Keywords: Time-gated fluorescence spectroscopy; autofluorescence; classification; fluorescent plastics labeling; hyperspectral imaging; image processing; plastics sorting and recycling.