Sagitto uses machine vision to measure colour
Authentication

From Bee Or Not From Bee

July 1, 2025
  •  
3 minute read
George Hill
Sagitto Ltd

This blog post explores some of the methods that can be used for testing honey for adulteration with syrups. We start with NIR analysis, then look at the C4 Sugars test, Spatial Offset Raman Spectroscopy (SORS), and finally discuss Nuclear Magnetic Resonance Spectroscopy (NMR). Machine learning can make an important contribution to all these techniques.

Some Background

Sagitto first became involved with honey analysis in 2015, when we applied machine learning to metabolomic data to create a set of unique biochemical ‘fingerprint’ markers to clearly distinguish nectars and honeys by floral type1.

The explosive interest in New Zealand manuka honey as a 'superfood' has led to very high prices for genuine manuka honey (as shown in this graph). However high price differentials for mono-floral honey also increases the economic incentive for adulteration - not only with cheaper honeys, but also with non-honey syrup substitutes such as corn and rice syrup. In response to this threat, over the past ten years a number of different analytical techniques have been developed for determining the floral origin of honey, and its potential adulteration. This blog post will discuss some of these.

Testing Honey Using NIR Spectroscopy

We first wanted to see whether NIR spectroscopy could be used detect honey adulteration. We obtained 12 New Zealand honey samples of known providence, and  scanned each with a Sagitto miniature NIR instrument to obtain its NIR absorbance spectrum.

Twelve New Zealand honeys
12 New Zealand honeys
NIR absorbance spectrum for New Zealand honeys
NIR absorbance spectrum for New Zealand honeys

We purchased corn syrup and brown rice malt syrup from a local supermarket, and mixed varying quantities with approximately 15gm of individual honeys to emulate adulteration of honey with cheaper syrups.

Corn syrup and rice syrup used as honey adulterants in our experiment

After scanning various mixtures of honey and syrup, we created two regression models - for clover honey adulterated with corn syrup, and mutlifloral honey adulterated with rice syrup.

Clover honey adulterated with corn syrup
Multifloral honey adulterated with brown rice malt syrup

These two regression models looked very promising. However when we generalised our model to include adulteration of other honeys, we noticed that the model tended to predict the presence of adulterant even in pure honeys (as circled in the lot below.)

New Zealand honeys adulterated with either corn or rice syrup

A classification model, that sought to distinguish between pure and adulterated honeys, showed a similar pattern of predicting false positives. Almost all adulterated honeys were correctly predicted, but 7 out of 12 pure honeys were incorrectly classed as adulterated. This result is consistent with a study2 done in 2014/15 by PerkinElmer using FT-NIR, and highlights a limitation in using NIR for the purposes of detecting honey adulteration.

7 out of 12 pure honeys were incorrectly classed as adulterated

C4 Sugars Screening

(to be completed)

Acknowledgements

Almost all honey used in this study was from two reputable and long-established family companies - Generation Honey and Airborne Honey. This table lists these honeys, together with their Brix and colour measurements.

References

  1. Unpublished research conducted in conjunction with the University of Waikato and Analyical Laboratories Ltd.
  2. Detection of Honey Adulteration Using FT-NIR Spectroscopy
  3. C4 Sugars Screening Test
  4. Application of Spatial Offset Raman Spectroscopy (SORS) and Machine Learning for Sugar Syrup Adulteration Detection in UK Honey
  5. Differentiation and classification of Australian single botanical origin honeys using spatially offset Raman spectroscopy
  6. Example NMR Honey Profiling Report

Subscribe to Sagitto's Blog

Get industry insights that you won't delete, straight in your inbox.
We use contact information you provide to us to contact you about our relevant content, products, and services. You may unsubscribe from these communications at any time. For information, check out our Privacy Policy.
George Hill
Sagitto Ltd
Sagitto's founder, George Hill, first started working with artificial intelligence during the 1980s, while developing 'expert systems' within Bank of America in London. On returning to New Zealand, he undertook part-time study with the University of Waikato's Machine Learning Group while working for Hill Laboratories, a well-known New Zealand commercial testing laboratory. This led to the formation of Sagitto Limited, dedicated to combining the power of artificial intelligence and machine learning with spectroscopy.

More news

Food

Just Peanuts?

Peanuts are a hugely valuable food crop, with worldwide production of more than 50 million tonnes. This blog post looks at how NIR spectroscopy can help peanut breeders and manufacturers of peanut products. We focus on what NIR can tell us about the composition of peanut kernels and peanut butter; and as an example we use NIR to measure sucrose when added to peanut butter.

Read Article
Data Science

Learning From Durum Flour

Good pasta is made from durum wheat flour rather than the softer wheat varieties used in bread. However being more expensive, durum flour ('semola rimacinata') is sometimes adulterated with other wheat flour. For a bit of fun, we've created a model that has the potential to detect adulteration of durum wheat flour. And we use this to demonstrate how Sagitto evaluates calibration models.

Read Article
Machine Vision

More Than Meets The Eye

We often make decisions based on colour. For example, we know that yellow bananas are riper than green bananas and therefore we expect them to be sweeter. But what if we could also see beyond the visible spectrum? Could we make even better decisions by combining visible image data with near infrared data?

Read Article