Enabling Continuous VNIR + NIR Analysis with FX Fusion

Adept Turnkey in collaboration with Australian Scientific Researchers

Food authenticity a true challenge faced by global food systems

Food adulteration remains difficult to detect and control. Two challenges are particularly persistent. The first is the detection of undeclared and potentially hazardous substances. The second is verifying attributes such as animal welfare, fair trade, and environmentally responsible production¹. While these characteristics are often considered beyond the scope of traditional analytical chemistry, their effects on the chemical composition of food products may still be observed and quantified.

Hyperspectral imaging has proven to be a powerful tool for food analysis because it captures detailed spectral information without destroying the sample¹,². However, until recently, practical limitations in sensor technology have restricted how much of the available spectral information could be used reliably.

There is quite a lot of work being undertaken in Australia by numerous organisations. Daniel Cozzolino a Principal Research Fellow at The University of Queensland and his colleagues have been studying meat from different species for some time. They have published a number of papers and continue to look for solutions to the problem.

Other scientific research efforts in Australia have investigated the use of hyperspectral imaging as part of a broad effort to explore meat provenance using the FX Fusion system designed by Adept Turnkey. For example research has been done to examine whether spectral signatures present in beef steaks could be linked to upstream information such as farm or carcass level data.

Meat authenticity and traceability are especially important within this context. Incidents involving non-declared species, such as horse meat substitution, have highlighted the need for reliable and scalable verification methods⁴. Beyond safety concerns, health and religion also play a major role in driving consumer expectations around meat authenticity¹.

FX Fusion: continuous hyperspectral data from 400–1700 nm

FX Fusion is the only dual camera hyperspectral system available commercially that fuses data at the pixel level from two hyperspectral cameras. In the case of some of the research done in Australia, the FX Fusion system combined visible near infrared and near infrared data into a single, pixel-aligned dataset spanning 400 to 1700 nanometres.

By combining data from two hyperspectral cameras into one continuous datacube, FX Fusion enables richer analysis and more spatial precision without the complexity traditionally associated with multi-sensor workflows.

The Managing Director at Adept Turnkey, Marc Fimeri commented that “Hyperspectral camera technology that combines two sensors with different wavelength ranges into a single camera is inherently difficult to build. It is expensive and prone to degradation with temperature and other environmental factors. It also forces the combined resolution to be that of the lower resolution sensor. Processing data after capture allows us to avoid these issues and to produce a very cost-effective method of combining wavelength bands accurately at the pixel level. We saw this gap in the market and filled it with FX Fusion.”

For applications such as meat provenance and authenticity, this capability is critical. Subtle spectral features can carry meaningful information about composition, processing, and origin. FX Fusion ensures those features can be analysed together with confidence.

Why spectroscopy works in meat analysis

Spectroscopy has long been used for food analysis, with near infrared techniques widely applied alongside chemometric and machine learning methods such as principal component analysis (PCA), partial least squares (PLS), artificial neural networks (ANN), linear discriminant analysis (LDA), and support vector machines (SVM)¹.

Near infrared wavelengths between 900 and 1700 nanometres contain information that is not visible to the human eye and often reveal spectral features associated with fat, moisture, and other compositional attributes relevant to meat¹. At the same time, visible near infrared wavelengths between 400 and 1000 nanometres capture colour and surface-related information that can also support authentication and discrimination¹.

When used together, VNIR and NIR data often provide higher reliability than either spectral range alone. Continuous spectral coverage allows complementary features to be analysed as part of a single workflow¹.

Meat provenance research in practice

In the work done by researchers in Australia, approximately one hundred beef steaks of different cuts were sourced from different butchers. VNIR (400–1000 nm) and NIR (900–1700 nm) hyperspectral data were captured simultaneously using the FX Fusion system.

After fusion, FX Fusion produces perfectly aligned reflectance datasets, combining Specim FX10 and FX17 data into a unified and precise output ready for in-depth analysis.

“Kevin Dowsey our Technical Director at Adept Turnkey and his team of vision engineers did a great job” Marc Fimeri said. Data produced by FX Fusion was sent to Specim, a leading worldwide manufacturer of hyperspectral cameras for review. Rainer Bars an hyperspectral imaging specialist and senior field engineer at Specim reviewed the data and commented: “The FX Fusion data is very good! The resampling of the FX17 data to FX10 spatial resolution looks great, with no interpolation artefacts."

Early results in these studies are described as promising, with identifiable spectral signatures observed within the fused datasets. The work demonstrated the potential of fused hyperspectral data for provenance-related research.

Combining VNIR and NIR data is technically complex

Although the benefits of combined spectral analysis are well understood, achieving reliable fusion is technically complex. VNIR and NIR hyperspectral cameras operate with different spatial resolutions, optical characteristics, and calibration requirements. Even small misalignments can undermine confidence in classification and modelling.

Without precise geometric alignment, VNIR and NIR information cannot be reliably correlated at the pixel level. This often results in extended calibration and post-processing workflows and reduced repeatability across projects.

FX Fusion addresses these challenges through a software-driven approach to hyperspectral data fusion. It performs precise geometric registration between camera models and outputs a single hyperspectral data cube with sub-pixel accuracy.

Fat content as a key driver of spectral variation

Fat content plays a central role in meat quality and is a major contributor to spectral variation¹. Fat influences reflectance across both visible and near infrared wavelengths, making combined spectral analysis particularly valuable.

Hyperspectral imaging allows fat-related information to be analysed spatially across a steak rather than as a single bulk measurement. This spatially resolved perspective provides richer insight into heterogeneity within samples and supports more informative modelling. For example, with high quality meat and a consumer trend towards marbled steak, fat is distributed in very small pockets throughout the meat. The ability to fuse data accurately to the sub-pixel level allows FX Fusion to accurately measure and determine the distribution of these small pockets of fat across the whole steak.

Broader implications in sustainability and trust

Advanced non-invasive technologies for meat authentication align with broader goals around responsible consumption and supply chain transparency. Meat fraud contributes to inefficiency and waste, amplifying the environmental impact of an already resource-intensive industry.

By supporting more reliable and scalable authentication methods, hyperspectral imaging and data fusion technologies can play a role in improving food system resilience. As climate change places increasing pressure on food security, tools that enhance trust and efficiency become even more important.

The future for FX Fusion

As food authentication challenges continue to evolve, hyperspectral data fusion provides a pathway toward more confident, efficient, and trustworthy analysis.

There are many other applications that can benefit from fused hyperspectral data, and Adept Turnkey has committed to ongoing development across different hyperspectral camera platforms, with future work potentially extending to fusion with additional imaging modalities.

References

1. Pischetsrieder, M. Global food-related challenges: What chemistry has achieved and what remains to be done. Angewandte Chemie International Edition, 57(36), 11476–11477 (2018).
2. Nolasco-Perez, I. M. et al. Comparison of rapid techniques for classification of ground meat. Biosystems Engineering, 183, 151–159 (2019).
3. EU Rapid Alert System for Food and Feed (RASFF). Horsemeat shipments destined for EU markets (2020).
4. Sridhar, A. et al. Advanced authentication technologies and sustainability in food systems (2023).