Detection and classification of heterogeneous materials as well as small particles using NIR-spectroscopy by validation of algorithms

Detection, characterization and sorting of plastics using Near-Infrared (NIR) Spectroscopy is State of the Art for plastic recycling processes. NIR active materials could be characterized according to the specific spectra of each material in the NIR spectrum. This works well for homogeneous materials, as they have known uniform spectra. For the detection of heterogeneous material and particles with smaller size than the resolution of NIR camera, the analysis becomes difficult due to mixed spectra. In this paper, the capturing of spectra information with a NIR camera takes place simultaneously with the optical detection with a VIS (Visible light) camera. The NIR spectra and VIS information are combined for the analysis of the mixed spectra, because the higher resolution of VIS camera contributes to a clear definition between the plastic materials and the background as well as between the selected materials. In order to determine the material composition, different kinds of mixed spectra of plastics with background as well as of plastic composition were detected and analyzed. The background is a conveyer belt made of black plastic, and the studied types of plastics are Polypropylene (PP), Polystyrene (PS), High-Density-Polyethylene (HDPE) and Polyvinylchloride (PVC). For the analysis, several algorithms will be developed and tested. In the end, a universal algorithm which performs well for all kinds of mixed spectra will be selected and improved.