Hyperspectral Imaging in the Recycling Industry: A Review
Updated: Jun 24, 2021
The recycling industry in the United States largely owes itself to the position we found ourselves in during World War II. Strapped for materials, the U.S. Army was struggling to provide the materials necessary for their overseas activity. With the foresight that sourcing these materials from scratch was prohibitive, a different approach was tried.
A national campaign was started to gather these materials from where they knew they already existed – the public. Newspaper ads and posters called for the donation of rubber, steel, and paper. Ultimately they sourced thousands of tons of these materials, and the recycling industry in America was born.
In the 1960’s the first curbside collection of recyclables began, with metal, glass, plastic, and paper products being the major categories. Some 50 years later this process continues, and in fact has expanded, with recycling technology increasing the types of materials that can be recycled and industry, in turn, increases the production of materials that can be recycled. Together these efforts have had dramatic effects on the production of raw materials that comprise them and the environment as a whole.
The most common form of recycling is called single-stream or one-sort recycling. When you see a blue bin by someone’s driveway that’s often what you’re seeing. It requires no separation on the part of the consumer. They simply place all compatible materials (which are often labeled on product containers) into the bin, and either bring it to a collection point or have it picked up.
This approach is effective in the sense that it encourages more participation in recycling by being simple and easy to do. But it’s not without its drawbacks. For instance,
-inclusion of non-recyclable materials
-presence of contaminants
To process these materials, and to address these problems, recycled material is sent to a Materials Recovery Facility (MRF). Here the recyclable material is separated from the non-recyclable material. The process is conducted in a series of steps:
1) Material is loaded onto conveyor belts
2) Employees remove non-recyclable material by hand
3) Screens separate material by weight
4) Magnets remove metal from heavier material
5) Workers sort paper products by hand
6) Quality control to ensure correct sorting and end-products
This process has historically been largely conducted by manual labour. That was problematic for three reasons
1) It’s a relatively dangerous activity
2) It’s expensive to do this all by hand
3) Human labour produces human error
But given a lack of alternative options, this was, until recently, the best option available. That has slowly changed in the last few decades with the addition of colour sorting. This is the process of using machine vision to identify recyclables and contaminants, and feeding that information into diversion and removal apparatuses to sort material. Colour sorting of recyclables has a few advantages and disadvantages.
- Quick & Cheap
- Not 100% effective
- Can’t distinguish between materials of the same colour
Despite these disadvantages, the idea of automated sorting of recyclables is a good one. The problem lies in its ability to monitor parameters outside of the visible range (e.g. other than color). This is where hyperspectral imaging has a lot to offer.
From a cursory point of view sorting with hyperspectral imaging looks nearly identical to color sorting. There’s a camera of sorts on the conveyor, it collects data, and that data is used to sort the material. The difference is the ‘camera’. With hyperspectral imaging, you aren’t limited to the visible spectrum. Instead, you can opt to use wavelengths in the UV, NIR, or IR regions. Additionally, hyperspectral imaging systems take spectral data at these wavelengths for each and every pixel of images taken. These two attributes allow for
-The detection of chemical hazards
-Distinguishing between materials of the same color
-2D distribution data
Taken together, the benefits of hyperspectral imaging allow for the effective automatic sorting of single-stream recycling. It allows for the removal of human labour and increases the effectiveness of automatic sorting. Instead of being inferior to manual separation, hyperspectral imaging offers advantages that humans cannot. For all these reasons it’s incumbent on recycling operations to upgrade to hyperspectral imaging-based automated sorting.