Holy Grail 2.0: digital water marking for automated plastic sorting

Plastics have gone from being the wonder material in the 1980s to a liability now – poor degradability, low recycling, plastic bags choking sewage systems, micro-plastics entering our foodchain through aquatic animals are just some of the problems that we associate with plastic. Infact there is a large stretch of ocean known as the Great Pacific Garbage Patch, that stretches from the West Coast?of North America all the way to Japan. But the ease of use that plastic offers is unique. It is lightweight, waterproof & cost effective, so wishing away plastics is unrealistic. The challenge, rather, is to collect & recycle so that the rate of generation of plastic waste goes down substantially over time. Countries have tried doing their bit through regulations like increasing the minimum thickness of plastic bags, making it mandatory for retailers to charge for them, banning single use plastic for trays & spoons and printing the rate per kilo that rag pickers and garbage collectors can get by selling it to recyclers or aggregators. While the extent of compliance varies across the country, the fact that industries and brands have to mandatorily balance books on generation and collection of plastic waste ensures certain level of accountability. Going forward, a certain proportion of recycled content could be made ?mandatory, like in the US, PET bottles for non-food applications need to have atleast 10% R-PET. ?

The good thing about plastic manufacturing is, being an extrusion process, it is quite accommodative of rework. Polymer chips – virgin or recycled, melt as they pass through the melt-extrusion process & the molten liquid is then blowing to form films or moulded into bottles. However, it is essential that that polymers are not mixed while recycling & to make it successful at scale, we need technology which can sort & separate different grades of plastic online at high speed, accurately. If you reflect on the types of shapes even for the same type of plastic that the system is likely to encounter across brands, products & sku sizes, the challenge increases manifold. Current sorting is done mainly NIR (Near Infra Red) – which determines the grade of polymer from the frequencies it absorbs when exposed to a wide band of IR frequencies. It is fast & accurate but it only works on transparent plastics & hence has to be complemented by manual sorting.

Holy Grail started as a P&G project in 2016 – but it must have been clear to the leadership at that time that a project of this nature would only succeed if there was buy-in from the entire value chain from retailers, packaging convertors, aggregators, recyclers, machine manufacturers as well as regulators. More importantly, it needed a large number of brands to come together to collaborate so that it would, over time become the industry standard – a bit like, Microsoft Windows. The original project sponsors were FMCG (P&G, Nestle, Pepsi, Danone, Henkel), Specialty chemical (BASF), Retail (Carrefour), Packing convertor (Constantia), Recycler (Suez), Machine manufacturer (Pellence ST) & Digital Watermark technology supplier Digimarc to name a few but it tells you the kind of broad-based consensus that it was able to achieve while it was still a pilot project. The first phase of the project between 2016-2019 was focused on finalizing the technology route & broadly two routes were found to be feasible. One was to incorporate a chemical marker in the plastic – a different one for each grade which could be detected in UV light at the recycling facilities. The other one was more interesting – digital watermarking.

Watermarking is a standard anti-counterfeiting tool – something which ensures that rip-offs can be easily detected – but how does one incorporate that into the artwork or substrate without taking away from the design & the brand is the question. Plus it also needs to be detectable, distinguishable and easily to incorporate. To understand digital watermarking, I needed a refresh of digital image processing. I am adding that to give the context and then we will see how this technology can be leveraged to obtain a digital footprint.

A digital photograph is made of dots or pixels. Higher the pixel concentration per unit area of the picture, sharper it is – expressed as its resolution. That is the difference between a normal & an HD feed on TV. Pixels use binary information – so if it is a black and white picture you need just two modes, on and off – characterized by 1 & 0. But even a black & white picture has various shades of grey & to represent it, the pixels need to exhibit different levels of intensity. This is achieved through bits - if it is an 8-bit picture, it means that each pixel is defined by a series of 8 binary digits from which could be anything like 00000000, 01000001, 11001001 & so on. Imagine that light from 8 different sources converge to form a beam. If all of them are switched off, there is total darkness. Depending upon the number of bulbs are glowing, the intensity will vary. There are 256 unique 8 digit numbers using only the digits 0 & 1 & each of these combinations represents a shade of grey. So by knowing the binary code fed to the pixel we can determine the colour shade.

This becomes a little more complicated when we consider colours. Visible light that we best know as the colours of the rainbow may seem like seven colours – but actually there are only three - red, green and blue and all other colours are combinations of these. To represent a colour image, the pixel therefore needs not one, but three sources of light – one corresponding to each of the basic colours. Each source of light will have multiple, say 8 bits. So we may know magenta is a combination of blue and red light but it is not one colour, but a superset of over 50s shades. While blue & red dominate, each colour is only completely defined when all the three, R-G-B values are specified. ??

Take a look at the picture below. I selected a colour at random and noted down its RGB values – which incidentally was 219/069/235 respectively. Then I tweaked the values, one by one & derived variations A, B & C– which we will call ‘modified tiles’ – which you will probably agree, look practically identical.

While printing the label or laminate, the modified tiles can be incorporated in the design multiple times using a grid as shown in the picture. It is a vitual grid but it ensures that the design occurs at a predefined frequency in a chosen location within the artwork. This ensures that the modified tile inevitably gets captured by the camera irrespective of the orientation of the bottle or label when it is being sensed. To the naked eye it seems like the same colour – but the camera uses detection software which not only identifies the presence of the modified tiles - by comparing it with the original code, identifies it as A, B or C & therefore the grade of plastic used! As visual inspection is generally very fast, this method ensures high throughput, accurate sorting. ?

A slight variation is used for bottles – wherein the moulds used for shaping the bottle have a characteristic texture imprinted in the shell which gets transferred to the bottle during the forming process. It is at microscopic level and texturual differences can be picked up only by cameras, but through the embossing, the material type is defined and therefore detected at the sorting station.

While digital water marking was originally conceptualized for material detection, as the technology matures to a point where the digital watermark can be detected & decoded using software available on smartphones, it will become a seamless way for brands to communicate with consumers, in addition to making it easier to segregate plastic at source by type – which in a way is the holy grail of plastic recycling.

Holy Grail 2.0 is about commercializing the technology and the team had a governance body with most of the initial backers and a few new ones like Amazon. It has its own Secretariat, facilitated by the AIM (European Brands Association) which is in charge of Project Management & is backed by a legal and advisory team to ensure that policy inputs and country rules are incorporated into planning & execution. The technology has been proven at semi-commercial scale to achieve 99% detection and 95% ejection across different types and pack formats with a throughput of 2.5 TPH for bottles and 500 kg/hr for flexibles like pouches and wrappers.

In his book Tipping Point, Malcolm Gladwell writes how an emerging trend is picked up by early adopters & as their tribe increases, it achieves critical mass or in essence reaches a ‘tipping point’ when a sudden discontinuity happens. With HG2.0 commercial rollout in Denmark, France & Germany this year, we may be at the threshold of such a tipping point. Who knows, in a few years’ time this technology may become so ubiquitous that we will forget that there was a time when we would struggle to separate plastic grades!