Artificial intelligence has learned to distinguish between 12 types of plastic

Scientists from Aarhus University in Denmark have createdPlastic classification via in-line hyperspectral camera analysis and unsupervised machine learning new technology that allows you to effectively sort plastic. The artificial intelligence model is faster and more accurate than existing counterparts.

Today, much of the plastic in recycling plants is still sorted by hand. This is a slow and time-consuming process with low pay for performers, who, moreover, often make mistakes.

In the same plants where sorting is automated, different types of plastics are usually separated from each other by density. But this approach does not work well for heterogeneous waste, so you have to use infrared spectroscopy. However, this increases the cost of processing.

In the new work, scientists from Denmark used hyperspectral data in the infrared range. The photos taken were analyzed by an artificial intelligence system. She was trained to distinguish signals corresponding to the most common types of plastic raw materials, for example, PET (polyethylene terephthalate), PVC (polyvinyl chloride), polyethylene, polypropylene, polystyrene.

A working prototype of the system was successfully tested: the sorting accuracy reached 96%. Spring 2022 technology will introduce at Danish waste treatment plants owned by PLASTIX and Dansk Affaldsminimering Aps.