A sensor fusion-based classification system for thermoplastic recycling

The first step in thermoplastic recycling is the classification of plastic wastes into categories. Nowadays, it is a manual task characterized by its high cost and inefficiency. This work presents an automated classification system for thermoplastics that can properly sort the most common automotive industry plastics. The core of this new automated system is that it fuses the information of three different sensors: a CCD of visible spectra, a NIR hyper-spectral sensor and inductive sensors. It is based on neural networks and has shown, through experimental validation, a very high accuracy (near 100% for certain thermoplastics). The proposed methodology is being integrated on an industrial prototype that further will be placed in a plastic recycling plant.