Blog
Hyperspectral cameras let sorting systems "see" the chemical fingerprint of every material on the belt – here's how it works and where it's used.
Walk past a modern optical sorting line and you might notice rows of cameras and sensors mounted above a fast-moving conveyor belt, scanning a constant stream of mixed waste. Behind that scene is one of the most important advances in recycling technology of the past two decades: hyperspectral imaging. Unlike a standard camera, which captures an image in just a few colour channels, a hyperspectral sensor captures dozens or even hundreds of narrow spectral bands across and beyond the visible spectrum – effectively giving each pixel its own detailed light "signature".
Every material reflects and absorbs light differently depending on its molecular structure. Plastics, paper, glass, metals and textiles each have a distinctive pattern of absorption across the near-infrared and visible spectrum – a kind of optical fingerprint. A hyperspectral camera records this fingerprint for every point on the conveyor belt, generating a data cube that combines spatial information (where an object is) with spectral information (what it is made of). Specialised software then analyses this data in real time, comparing each material's spectral signature against a library of known references to classify it.
Hyperspectral imaging allows facilities to sort materials that look visually identical to the human eye but are chemically very different – for example, separating PET from PLA bioplastics, or identifying which fraction of a mixed plastic stream contains PVC, which can contaminate entire batches of recycled material if even small amounts are present. Because the technology operates at conveyor speed and across the full width of the belt, it scales far beyond what manual sorters could ever achieve, while maintaining a level of material-level precision that was simply not possible with earlier colour-based or single-wavelength sensors.
Beyond recycling, hyperspectral imaging is used across agriculture, food safety, pharmaceuticals and remote sensing – anywhere that distinguishing materials by their chemical composition rather than their appearance adds value. In each case, the same underlying principle applies: capturing a detailed spectral signature for every point in an image and using it to classify what is being looked at.
In waste sorting, once a material is classified, the system triggers an array of precisely timed air jets or mechanical actuators that eject the targeted material from the stream into the correct collection channel, all within a fraction of a second. For textiles, hyperspectral sensors can detect fibre blends such as cotton-polyester mixes, which is critical information for deciding the best recycling route for each garment. This combination of speed and accuracy is what allows modern Material Recovery Facilities to produce the high-purity output streams that recyclers and manufacturers demand.
Hyperspectral systems generate large volumes of data that must be processed in real time, and classification accuracy depends on having well-maintained reference libraries for the materials being sorted. PICVISA integrates hyperspectral and multi-sensor imaging across its optical sorting range, combining this detailed material data with artificial intelligence and deep learning to continuously refine sorting decisions – equipment that adapts to variations in lighting, contamination and material condition, delivering consistent purity even as the composition of incoming waste streams changes from one batch to the next.
Discover our optical sorting technologyGet in touch with our team to discover how PICVISA's optical sorting and robotics solutions can fit your recycling operation.