Efficient recycling of waste into reusable raw materials is one of the significant efforts we must take to stop global warming and the over-exploitation of natural resources.

The environmental benefits of recycling are clear. Recycling conserves natural resources and reduces greenhouse gases and pollution, and the use of fossil fuels in energy production. It reduces energy consumption by about 70 % for plastics, 60 % for steel, 40 % for paper, and 30 % for glass.

A significant value lies in reusable material. However, we are still far away from our recycling targets. Most of the collected waste is still used for energy production and burnt in power plants – not reused. Price is often a factor in low recycling rates, as it is often cheaper to produce new products from raw materials than recycled materials.

To make recycling not only ecologically but also economically viable, reusing materials needs to be cheaper and easier than using virgin materials. With proper material handling methods, different materials can be efficiently recycled and turned into profit. This is where hyperspectral imaging can make a difference.


A typical waste management process includes the collection of waste in a recovery facility, segregation into different waste fractions, cleaning, and final classification into materials that are placed in landfills, burned, or recycled based on the type and purity.

The sorting process is a critical step in recycling. Better sorting accuracy means better separation of different grades of material, which results in higher value recovery. A typical sorting process is based on a mix of techniques and cannot rely on just one detection technology. The detection technology used often limits the types and the amount of the collected material that can be sorted.

Most of the recycling plants use different technologies from bar code readers and RGB cameras to X-Ray and Eddy current systems. While they are capable technologies to a certain extent, they are not perfect solutions as their capability to identify the material is limited.

For example, if a plastic bottle is missing the barcode, it is not possible to detect if it is PET or HDPE. Eddy-current detectors can sort out conductive metals but not separate plastics or pulp. RGB cameras can sort bottles into transparent, black, and colored but cannot distinguish one plastic type from another.

When the recycled portion is not pure enough for reuse, we lose recyclable material to landfill or energy production. The poor sorting result also results in lost profits, which makes recycling unprofitable and dependent on public support.

Different waste streams require different detection and processing methods to be recycled efficiently and current recycling methods are not flexible, efficient, and informative enough to tackle the challenge.

To make up for inadequate detection technologies, human labor is still used. Sorting waste by hand is slow, inaccurate, expensive, and dangerous, and separating different plastic types from each other remains impossible because the human eye cannot tell them apart.

To work efficiently, profitably, and safely recycling plants must have sensors capable of separating different materials reliably and with high purity. Hyperpectral imaging offers a powerful technology for accurate and sustainable waste recycling.


Hyperspectral cameras can differentiate materials accurately and reliably based on their chemical composition. They measure and analyze the spectrum of light reflected from or transmitted through the material. When measuring the spectrum beyond the visible region called near-infrared (NIR), we see that chemically different materials have unique spectra.

Multispectral technology has improved the situation; however, it has its limitations. Multispectral cameras acquire spectral data typically with one to three, or in some cameras, a maximum of 8 spectral bands, meaning that in each sorting location, it identifies only a few basic materials. The purity of the result is also often limited as there are interfering factors in the material stream. (Read more about the difference between multispectral and hyperspectral cameras)

The use of hyperspectral imaging in waste sorting has been restricted by the insufficient performance of hyperspectral cameras in terms of speed, spatial resolution, ruggedness, connectivity, and high cost – until recent years.

The recent development has improved both speed and resolution of hyperspectral cameras, while their implementation cost now meets the ROI criteria of commercial solutions. Furthermore, the algorithms and solutions for the real-time processing of a large amount of data produced by hyperspectral cameras are now available.

For in-line sorting applications, a line-scan hyperspectral camera is the only practical and properly working solution, as it captures the entire spectral data of the full material stream from each pixel in the line precisely at the same time with a single scan.

A line-scan (push-broom) hyperspectral camera can be installed on existing and new sorting lines with proper illumination and a real-time data processing solution like any line-scan camera. The material identification result, pixel by pixel, is available through a standard interface to commercial machine vision systems. The results can then be used to control the air nozzles or picking robots.

A hyperspectral camera solution provides superior performance and several benefits in various waste treatment processes over conventional sensor technologies, as summarized in Table 1.

Table 1. Added value by hyperspectral imaging in sorting different types of waste streams

When used together with other technologies, hyperspectral cameras increase sorting accuracy by providing precise information on material type. The latest generation of hyperspectral cameras can increase the purity of recycled materials by close to 100 %. Increasing the purity of recycled plastic by even a few percent can double its value. Extracting more recyclable material also means that we are disposing of less waste in landfills.

Compared to a multi-spectral camera with fixed spectral bands, the hyperspectral camera is flexible and can adapt to sorting various waste streams. It can also adopt new sorting algorithms when they become available.


Out of all the plastic manufactured, only 9% gets recycled. 12% is incinerated for energy, and 79% goes to landfills or nature. It is estimated that by 2050 there will be more plastic in the oceans than fish. The majority of non-recyclable plastic waste comes from not being able to separate different plastic types from each other reliably.

When we sort and separate plastic, high-quality and valuable polymers can be reused. The main objective in sorting is to reduce the quantity of non-targeted plastic polymers and the number of non-plastics like paper, metal, glass, oil, soil, or other contaminants. There may also be unwanted additives like flame retardants within the plastic, that can be detected, identified, and sorted with hyperspectral cameras.

Different polymers have identifiable spectral signatures in the NIR spectral region and can thus be sorted. However, many of the spectral signatures are close to each other. Here, the hyperspectral camera’s high spectral resolution is key to high sorting accuracy. With PP, PE, and PET plastics, for example, close to 99% purity can be achieved. (Read more: How Prodecologia used cutting-edge hyperspectral imaging technology to achieve 98 % polymer purity?)


A large fraction of recyclable plastic constitutes of black plastics, used especially in the automotive and electronics industries, which have added carbon-based pigment to produce the dark grey or black color. Black plastic types have been notoriously difficult to identify, and so far, there has been no reliable sensor technique to sort these materials for reuse. Even NIR hyperspectral cameras struggle, as the black carbon-based pigment absorbs practically all the NIR light.

In addition to the NIR region, different plastics have characteristic spectral features in the longer infrared region called mid-wave infrared (MWIR) where most black pigments are ‘less black’ (less absorbing) than in the NIR region. Thus, MWIR light can penetrate in and reflect from black materials, making their spectral identification possible.

With the Specim FX50 hyperspectral camera that operates on the MWIR region, we can sort black ABS plastics with close to 99 % purity. It is the only hyperspectral camera currently on the market operating in the MWIR region with the required speed, resolution, and sensitivity for industrial in-line use.

Below is an example of black plastic sorting measured in a laboratory with a Specim FX50 hyperspectral camera. Twelve pieces of ABS and PE were measured together with ten pieces of PS (34 altogether). For each sample group, half of the samples were shiny, and the second half with diffuse surfaces. The figure below shows that samples made of ABS, PS, and PE could be accurately sorted with the Specim FX50.


Textile recycling reduces environmental impact compared to incineration and landfilling. Nearly 100 percent of all textiles and clothing are somehow recyclable if they can be correctly classified and separated based on used fiber type.

One obstacle to increasing textile recycling has been the fact that various fibers that comprise clothing make reprocessing and recycling a challenge. Although it is possible to use human labor for classification, this is hardly economically feasible and poses a lot of error sources.

Hyperspectral cameras in the NIR spectral region can separate the most common types of textile fractions enabling automatic robotized processing. The NIR hyperspectral camera-based textile sorting has multiple benefits:

  • Non-contact and suitable to be applied in a conveyer belt
  • Gives information about both pure and mixed materials (qualitative and quantitative sorting)
  • Classification is not sensitive to used colors or dyes
  • Easily configurable for different sorting lines and new materials
  • For precise color information, hyperspectral camera can replace the RGB camera.
  • Cotton is an extremely resource-intense crop in terms of water, pesticides, and insecticides. Using recycled cotton can lead to significant savings of natural resources and reduce pollution from agriculture. Some materials such as cotton and linen can be recycled for car insulation or composted, but petroleum-based fibers such as polyester have little chance for reuse.


Although the material recycling percentage is increasing, it will still be necessary to incinerate some parts of non-recyclable material. These “waste-to-energy” power plants receive material from various sources like commerce, construction, household, and industry and use this for generating power in Refuse Derived Fuel (RDF) power plants.

The value of RDFs is derived from the calorific content (Image KK) – which is determined by material type. Certain materials like glass, rock, or dirt, have zero calorific value. Water content and ice will also affect the process.

Precise combustion process control and calorific values can only be calculated based on proper material recognition. Hyperspectral imaging in the NIR region provides an in-line solution for this.


Improved sorting accuracy increases the purity and value of the recycled material and the percentage of waste that can be reused. To improve sorting accuracy, we need better detection systems.

The potential impact of hyperspectral imaging on the recycling industry and society is significant. The hyperspectral camera is an accurate, reliable, non-destructive, and contactless detection tool that improves operational efficiency, enhances material purity, and improves profitability.

Advanced hyperspectral camera technology, analysis software, and spectral libraries are already available and in use in modern recycling machines and waste-sorting facilities and are expected to grow in the future due to the growing need to solve previously unfeasible sorting tasks.