---
title: "Multispectral cameras in food processing | Quantlix"
description: "A multispectral camera records light the eye cannot see and can reveal hidden bruises, foreign material and early spoilage. Learn how to test one on a line."
canonical: "https://quantlix.com/en/blog/multispectral-cameras-food-processing/"
last-updated: 2026-11-21
---

Source: https://quantlix.com/en/blog/multispectral-cameras-food-processing/
Language: en
Description: A multispectral camera records light the eye cannot see and can reveal hidden bruises, foreign material and early spoilage. Learn how to test one on a line.

Playbook Food manufacturing

# What can a multispectral camera see that a normal camera cannot?

A multispectral camera records light the eye cannot see and can reveal hidden bruises, foreign material and early spoilage. Learn how to test one on a line.

[Shahmir Abbas](https://quantlix.com/en/about/leadership/shahmir-abbas/)

Food Technology Lead

Nov 21, 2026 (2026-11-21) 14 min read

![A microchip held in tweezers above a green circuit board, with the headline “Seeing what the eye cannot”.](https://quantlix.com/_astro/insight-multispectral-cameras-food-processing.BKT9tN72_Z2dnWWH.webp)

01 / Engineering notes

Field notes: Blog

**On this page**

The short answer

A multispectral camera records light in more bands than the three colors that eyes and ordinary cameras use, including infrared. Water, fat, sugar and plastics reflect those bands differently, so in laboratory research spectral imaging has shown hidden bruises, foreign material and spoilage markers. It works best on one defined problem, tested on your own product.

Picture a scrap of clear plastic, the kind that tears off a liner bag, lying on a raw chicken fillet under factory lights. Pale, glossy and a little wet, it almost disappears against meat that is pale, glossy and a little wet. A person on a moving line can miss it, and so can an ordinary camera.

A **multispectral camera** records light in several separate bands, usually more than the three (red, green and blue) that an eye or an ordinary camera uses, and often including infrared. In some of those bands, plastic and meat can look nothing alike. This guide explains how that works, what research has found, where the limits are, and how to test one on your own product before you spend anything serious.

## How can a camera see what the eye cannot?

A camera can see more than an eye because it can record bands of light that people do not sense, and different materials reflect those bands differently. Two things that look identical in red, green and blue can look quite different in infrared.

Think of light as a very long piano keyboard. Human eyes pick up three broad notes: red, green and blue. An ordinary camera copies us and records the same three. A multispectral camera listens to more notes, and some of them sit beyond the red end of the keyboard, in the infrared, where our eyes hear nothing.

That would only be a curiosity if every material sounded the same in those extra notes. It does not. Water, fat, sugar, protein and plastics each respond to those bands in their own way, and a small computer beside the camera can learn the differences and flag an item. Damage can also show up before the eye can see it, as the research below shows.

## What is the difference between multispectral and hyperspectral cameras?

A multispectral camera records a handful of chosen bands, while a hyperspectral camera records hundreds of narrow ones. Suppliers draw the line in different places, so the useful questions are how many bands a camera records and which ones.

| Kind of camera | How much light it records | What it is good for |
| --- | --- | --- |
| Ordinary color camera | Three bands: red, green, blue | Anything a person can see: shape, color, size, labels |
| Multispectral camera | A handful of chosen bands, often including infrared | Fast checks for a known problem, such as a hidden defect or a foreign material |
| Hyperspectral camera | Hundreds of narrow bands | Finding out which bands matter, and detailed measurements |

The two spectral cameras work well as a pair. A hyperspectral camera is like a full blood test: it measures everything, which makes it slower and full of information you will not need every day. A multispectral camera is like the two or three readings a doctor checks at every visit, chosen because they answer the question.

The research shows the pattern. The apple camera described below recorded 288 wavebands. In a 2026 salmon study, the team used an algorithm to pick 30 wavelengths from a 400 to 1000 nanometer range and kept prediction accuracy high with far less computation. Explore with many bands first, then run the line on a few.

## What have researchers found with spectral imaging in food?

In laboratory studies, spectral imaging has found latent bruises in apples, small foreign pieces on chicken and spoilage markers in salmon. All three studies below used hyperspectral cameras, so they show what the light can reveal, not what a plant’s line has already achieved.

| Study | What was imaged | Result | Limit to remember |
| --- | --- | --- | --- |
| [Sensors, 2021](https://pmc.ncbi.nlm.nih.gov/articles/PMC8348186) | Latent bruises on 36 Golden Delicious apples, short-wave infrared, 930 to 2500 nm | Bruises that RGB images did not show became evident; the paper’s text reports detection accuracy from 89.6 to 100 percent | One variety, small sample |
| [USDA, Applied Sciences, December 2021](https://www.ars.usda.gov/research/publications/publication/?seqNo115=387652) | Foreign material on broiler breast meat, visible to short-wave infrared, 400 to 2500 nm | About 95 percent of 5 by 5 millimeter pieces in the training set (USDA’s summary), and 90.9 percent of 2 by 2 millimeter pieces in the test set | Laboratory samples |
| [Foods, February 2026](https://pmc.ncbi.nlm.nih.gov/articles/PMC12939719/) | Atlantic salmon, 400 to 1000 nm, two spoilage measures | A match of about 0.97 on a scale where 1 is perfect, against laboratory measurements | Laboratory study |

The apple study’s authors wrote that “RGB images showed no visible bruises,” while image reconstruction made them evident. The USDA study used thirty types of foreign material that could be found in poultry plants, with examples including plastic and metal equipment pieces, rubber gloves and pallets. It also reported no false detections in meat images without foreign material. The salmon team imaged the fish without cutting it open, which is the appeal: an estimate of the laboratory result for every item, with no sample and no wait.

## Is a spectral camera fast enough for a production line?

Spectral cameras are increasingly fast enough for production lines, though the best evidence for the fastest lines is still a simulation. They produce far more data than ordinary cameras, and for years that made them too slow.

In February 2025, researchers at the USDA Agricultural Research Service and the University of Georgia [reported](https://pmc.ncbi.nlm.nih.gov/articles/PMC11819658/) that compressing a deep-learning model and running it on a graphics processor (GPU) cut inference time (the time the model takes to analyze each image) by up to five times compared with unoptimized inference, and cut the model’s size by 50 percent, with comparable accuracy. Their simulation, based on two hypothetical line-scan cameras, indicated that the approach could keep pace with poultry lines running 140 to 250 birds a minute. It was not a plant trial. Running that kind of model on a small computer beside the line is covered in our article on [edge AI vision for food inspection](https://quantlix.com/en/blog/edge-ai-vision-food-inspection/).

Camera-based grading already runs at line speed. A 2021 [paper in the Journal of Food Science](https://pubmed.ncbi.nlm.nih.gov/34940977/) describes an in-line bell pepper system sorting about 3,000 peppers an hour into five classes by maturity and size, with reported accuracy of 96.9 to 98.7 percent. The abstract does not describe the camera, so read it as evidence for line speed, not for spectral imaging.

## Who builds spectral inspection for food lines?

Several specialist companies sell spectral cameras and sorters for food. Naming them is not an endorsement, and we describe only what each has said publicly.

- Specim. As of November 2026, its FX series cameras are, in the company’s words, “the first hyperspectral instruments designed specifically for industrial machine vision.” The [Specim FX page](https://specim.com/fx/) lists food quality as a use for the FX10 and food and feed quality for the FX17.
- Key Technology. In July 2025, Key Technology, part of Duravant, [introduced](https://www.key.net/en/news/key-technology-introduces-compass-belt-fed-optical-sorters/) a belt-fed COMPASS sorter for wet, sticky and delicate products, with up to 8 channels of multispectral sensor data. The company says that for many applications it eliminates the need for costly laser or hyperspectral sensors while keeping comparable sorting performance.
- P&P Optica. In July 2021, The Spoon [reported](https://thespoon.tech/pp-optica-raises-fresh-funding-for-hyperspectral-food-safety-inspection-tech/) a funding round for this Ontario company, whose hyperspectral system can detect foreign matter such as rubber, cardboard or plastic and is aimed mainly at meat products. That source is several years old, so check the current product.
- TOMRA Food. As of November 2026, its [nuts and dried fruit page](https://www.tomra.com/food/machines/nuts-and-dried-fruit) says its optical sorters identify and remove foreign materials and defective product, and it names X-ray and pulsed LED sensor options. Spectral cameras are one tool among several.

## How would you test one on your own product?

Test a spectral camera in five small steps, each of which can stop the project early if the answer is no. The example below is an illustration, not a real plant or a real result.

Say a poultry plant had a customer find a scrap of clear liner plastic in a pack of fillets. A metal detector cannot see plastic, and inspectors cannot reliably see it either.

1. **Name the problem precisely.** Not “foreign material” but “clear liner plastic, 5 millimeters and larger, on skinless breast fillets, on line two.” The narrower the question, the cheaper the answer.
2. **Build a sample box.** Fillets, plus scraps of every plastic that could plausibly reach the line: liner bags, glove fragments, strapping, label backing. Add the things that should *not* be flagged, such as fat, membrane and bone, because a system that cries wolf gets switched off.
3. **Scan the box.** A laboratory or a camera supplier images the samples with a hyperspectral camera and looks for the few bands where plastic and meat separate. The USDA study shows this can work in a lab; your plastic, meat and lighting decide whether it does for you.
4. **Trial on one line.** A multispectral camera tuned to those bands runs over the line for several weeks. A person checks every flagged item, so you learn what it catches and how often it raises a false alarm.
5. **Judge it against the numbers.** How many real pieces did it find? How many good fillets did it set aside? What would that rate mean over a year? Only then does a decision about other lines make sense.

## Where does a spectral camera pay off, and where does it fall short?

A spectral camera pays off where an ordinary one cannot do the job, and it falls short for anything deep inside a product. Use a spectral camera only for the problem an ordinary camera cannot solve, and test it on your own samples first.

Four places are worth testing:

- At receiving. Grading fruit and vegetables by ripeness and hidden damage as they arrive, so each batch goes to its best use.
- At sorting. Removing foreign material that is the same color as the product, such as shell in nuts or stems in frozen berries.
- During processing. Checking moisture, fat or sugar on every item instead of a laboratory sample taken once an hour.
- Before packing. Catching early spoilage, so the freshest stock can go on the longest journeys. We look at that in [how to extend food shelf life with new technology](https://quantlix.com/en/blog/extend-food-shelf-life-technology/).

The limits are simple to state:

- Depth: the cameras see the surface and a little below it, so a stone inside a date or a bone deep in a fillet needs another tool, usually X-ray.
- Lighting: they need steady, purpose-built lighting that is checked from time to time.
- Training: they must be taught on your product, and a model built on one variety may need adjusting for the next.
- Cost: they can cost more than ordinary cameras, so the sensible plants save them for the one or two problems that need them.

A camera flag is one check inside your food safety system, not a replacement for it. Your HACCP (hazard analysis and critical control points) plan, which our [HACCP guide](https://quantlix.com/en/standards/haccp/) explains, and your foreign-material controls define what must be checked and what happens to a flagged item. Your quality team decides on holds and releases.

For scale, the US Food and Drug Administration’s [compliance policy guide on hard or sharp foreign objects](https://www.fda.gov/regulatory-information/search-fda-guidance-documents/cpg-sec-555425-foods-adulteration-involving-hard-or-sharp-foreign-objects) (issued 1999, revised May 2005) treats objects of 7 to 25 millimeters in length as a potential hazard. Confirm which rules apply to your product and market with your food safety and regulatory advisers. For the wider picture, see our [overview of food industry challenges](https://quantlix.com/en/blog/food-industry-challenges-2027/).

## What should you ask a supplier before you buy?

Ask the supplier to prove the system on your product, not on a demonstration sample. These six questions are a good start:

- Will you scan our own samples, including the contaminants we actually find, before we commit?
- Which bands does the system use, and why those for our problem?
- How does it cope when the product changes: a new variety, a new season, wetter or drier stock?
- How fast was the line it was tested on, compared with ours?
- What happens to a flagged item, and how do we confirm it really left the line?
- How often does the lighting need checking or calibrating, and who does it?

A supplier who welcomes the first question is usually worth talking to.

## Where to start

Everything below can be done this month without buying anything.

1. Write the hidden problem in one sentence, with numbers if you can: how often it happens, where it is found, and what it cost the last time.
2. Start the sample box. Keep real examples of the problem and of the look-alikes that should not be flagged, labeled and dated.
3. Photograph samples under your own line lighting. If a person looking at the photos can tell the difference, an ordinary camera may already be enough.
4. Ask your quality team where a flag would fit in your HACCP plan and who would act on it.
5. Keep the six supplier questions for your first call with a specialist.

If you would like a second pair of eyes, tell us about one machine or one line and we will reply with a short note on what we would measure first. You can start on our [contact page](https://quantlix.com/en/contact/), and you can see what we offer producers on the [food and beverage page](https://quantlix.com/en/industries/food-and-beverage/).

## Key takeaways

1. Multispectral cameras record several light bands, often including infrared, so materials that look alike in color can look different to the camera.
2. The food results cited here (apple bruising, chicken foreign material, salmon freshness) are laboratory studies with hyperspectral cameras, not production lines.
3. Use a spectral camera only for a problem an ordinary camera cannot solve, and test it on your own samples before you buy.
4. A camera flag is one check inside your food safety plan, and your quality team still decides what happens to a flagged item.

## Questions people ask

### What is a multispectral camera?

A multispectral camera records light in several separate bands, often including infrared light that people cannot see. Because different materials reflect each band differently, it can tell apart things that look the same to a person or an ordinary camera, such as a foreign object on pale meat or a latent bruise on an apple.

### What is the difference between multispectral and hyperspectral imaging?

A multispectral camera records a handful of chosen bands, which keeps the data small enough for a fast line. A hyperspectral camera records hundreds of narrow bands, which gives far more detail but needs far more computing. A common pattern is to explore with a hyperspectral camera, then put a multispectral one on the line.

### Can a camera detect plastic or other foreign material in food?

In some cases, yes. A 2021 United States Department of Agriculture study found pieces of foreign material about 5 by 5 millimeters on chicken breast with roughly 95 percent detection in the training samples. Results depend on the material, the product and the lighting, so testing on your own samples matters more than any published figure.

### Can a multispectral camera see inside fruit?

Only a little way. These cameras read the surface and a thin layer beneath it, which is enough to reveal latent bruising or to estimate properties such as moisture. A pit, a stone or a defect deep in a thick product needs another method, and X-ray is the usual choice for dense objects.

### How is a multispectral camera different from X-ray inspection?

They answer different questions. X-ray looks through a product and is generally used for dense objects such as metal, glass and stone. A spectral camera looks at the surface and just below it and reads how materials reflect light, which can separate plastic from meat. Some plants may need both, each for the job it does best.

### Do I need a multispectral camera or will an ordinary camera do?

Start with an ordinary camera and good lighting, because it is cheaper and often enough for anything a person can see, such as size, shape, color and labels. Move to a spectral camera when the problem is invisible in color, such as same-color foreign material or hidden damage, and a test on your samples shows it separates them.

## Sources

1. [Nturambirwe, Perold and Opara, Classification learning of latent bruise damage to apples using shortwave infrared hyperspectral imaging, Sensors (2021)](https://pmc.ncbi.nlm.nih.gov/articles/PMC8348186)
2. [Chung and Yoon, USDA Agricultural Research Service, Detection of foreign materials on broiler breast meat using fusion of visible near-infrared and short-wave infrared hyperspectral imaging, Applied Sciences (16 December 2021)](https://www.ars.usda.gov/research/publications/publication/?seqNo115=387652)
3. [Jiang and others, non-destructive freshness assessment of Atlantic salmon with hyperspectral imaging and a Transformer model, Foods (February 2026)](https://pmc.ncbi.nlm.nih.gov/articles/PMC12939719/)
4. [Khan, Yoon and Bhandarkar, Deep learning model compression and hardware acceleration for high-performance foreign material detection on poultry meat using NIR hyperspectral imaging, Sensors (6 February 2025)](https://pmc.ncbi.nlm.nih.gov/articles/PMC11819658/)
5. [Mohi-Alden and others, Design and evaluation of an intelligent sorting system for bell pepper using deep convolutional neural networks, Journal of Food Science (23 December 2021)](https://pubmed.ncbi.nlm.nih.gov/34940977/)
6. [Specim, FX series hyperspectral cameras (product page, read 21 November 2026)](https://specim.com/fx/)
7. [Key Technology, Key Technology introduces COMPASS belt-fed optical sorters (15 July 2025)](https://www.key.net/en/news/key-technology-introduces-compass-belt-fed-optical-sorters/)
8. [TOMRA Food, nuts and dried fruit sorting and grading solutions (product page, read 21 November 2026)](https://www.tomra.com/food/machines/nuts-and-dried-fruit)
9. [The Spoon, P&P Optica raises fresh funding for hyperspectral food safety inspection tech (8 July 2021)](https://thespoon.tech/pp-optica-raises-fresh-funding-for-hyperspectral-food-safety-inspection-tech/)
10. [US Food and Drug Administration, CPG Sec. 555.425, Foods, adulteration involving hard or sharp foreign objects (issued 1999, revised May 2005)](https://www.fda.gov/regulatory-information/search-fda-guidance-documents/cpg-sec-555425-foods-adulteration-involving-hard-or-sharp-foreign-objects)

Sources checked on Nov 21, 2026.

- Playbook
- Food manufacturing

Written by

## Shahmir Abbas

Food Technology Lead · Quantlix

[See full profile](https://quantlix.com/en/about/leadership/shahmir-abbas/)

[Talk to Shahmir](https://quantlix.com/en/contact/)

[More blog](https://quantlix.com/en/blog/)

## Keep reading

[More blog](https://quantlix.com/en/blog/)

- ![A worker in gloves filling food trays on a conveyor, with the headline “Food industry challenges in 2027”.](https://quantlix.com/_astro/insight-food-industry-challenges-2027.vpPBxeT6_Z2qPeGf.webp)

  Perspective

  Food manufacturing Nov 21, 2026 (2026-11-21)

  ### [What are the biggest challenges in the food industry in 2027, and what solves them?](https://quantlix.com/en/blog/food-industry-challenges-2027/)

  Seven challenges food makers face in 2027, from spoilage and quality checks to downtime, skills, supply shocks and costs, and the tools that help with each.

  20 min read Read
- ![A row of automated machine cells with cable looms in a factory, with the headline “A camera that checks every item”.](https://quantlix.com/_astro/insight-edge-ai-vision-food-inspection.C95-UJkG_Z1OhuA9.webp)

  Playbook

  Food manufacturing Nov 21, 2026 (2026-11-21)

  ### [How can a camera check every item on a food line?](https://quantlix.com/en/blog/edge-ai-vision-food-inspection/)

  A camera and a small computer beside a food line can check every item all shift, even offline. Here is how edge AI vision works and how to start.

  15 min read Read
- ![Frosted plums and raspberries, with the headline “Fresher for longer”.](https://quantlix.com/_astro/insight-extend-food-shelf-life-technology.26TC7Fwa_VMc1L.webp)

  Playbook

  Food manufacturing Nov 21, 2026 (2026-11-21)

  ### [How can technology make food last longer?](https://quantlix.com/en/blog/extend-food-shelf-life-technology/)

  How technology makes food last longer: slow spoilage with steadier cold, read how much life each batch has left, and ship the shortest-lived stock first.

  14 min read Read

## Need the version tailored to your platform, risk model and roadmap?

Bring us the context. We will turn the patterns in this library into a pragmatic delivery path your team can use.

[Plan a workshop](https://quantlix.com/en/contact/)
