Smart algorithm for black plastic recycling

RESEARCHERS have developed a new system for sorting plastics more cheaply and at higher volumes, and opens up a new route to separating out black plastics.

Conventional sorting systems operate within the near-infrared range, which allows them to categorise plastics. However, this technique does not work for black plastics as the soot that gives them their dark colour absorbs most of the signal making them invisible to the optical system.

These black plastics are used in large volumes for car interior panelling and the urgency to recycle them is growing as the EU sets higher thresholds for car recycling programmes.

A team from the Fraunhofer Institutes in Wachtberg and Karlsruhe, Germany have developed a system that uses a radar detection camera and an intelligent computer algorithm that can sort through dark coloured plastics at much lower costs and with the similar degrees of accuracy compared to high frequency cameras currently used.

The team are using the system to filter assorted types of shredded plastic by depositing assorted plastic waste onto the conveyor belt where they fall off in a wide arc at a speed of 2–3 m/s. A radar camera module emits a beam of radiation, of a frequency in between infrared and microwaves, through this stream of falling flakes.

The system analyses how the individual particles have modified the radiation signals and determines what kind of plastic is present based on the spectra received; the colour is inconsequential.

If there are non-recyclable materials present, the system can decide to knock a particle out of the stream using a precisely aimed blast of air. A colour camera is also used to provide additional information about the shape of the object, in order to help the system open the blower nozzles at the right time.

The team says while the cameras they are using are less precise than the ones used in industry today, the addition of self-learning algorithms that detect minute differences in spectra has created a system suitably accurate that costs much less. The team say the system increases its accuracy over time - currently performing at 98–99% - at around the price of hyperspectral camera (€1,000–€10,000) compared to up to €1m (US$1.1m) for high frequency cameras currently in use.

Thomas Längle, head of optronics, system technologies and image exploitation (IOSB) at Fraunhofer, said: “For the first time, we have developed an affordable sorting system that detects every plastic colour, including black, both in real time and in large quantities.”

The team say the camera and algorithm solution is not limited to recycling applications. The modular design and the ability to expand the frequency detectors like lenses can also make it applicable to other conveyor belt systems that require sorting or inspecting, such as steel rollers and food production.

The team expect the camera to be available for recycling centres in early 2017 and will be ready for general market release by the end of that year.