VideometerLite+ Detects Chicken Freshness and Origin

Recently published in Sensors, a peer reviewed study evaluated whether a single portable multispectral imaging (MSI) device can assess multiple quality parameters in chicken at once: freshness, physical condition (fresh versus frozen and thawed), and geographic origin.

The study was carried out by the Agricultural University of Athens (AUA). It builds on the foundation of the TraceMyFish project, a collaboration involving AUA, the University of Iceland, Norway’s NTNU, Scio Systems, and Videometer. TraceMyFish focused on how spectral imaging could support traceability and quality control across food supply chains. This study extends that work from seafood to poultry.

Study Design

The researchers analyzed chicken breast fillets of Greek and Danish origin. Samples were tested in both fresh and frozen thawed condition, across multiple production batches.

Measurements were taken in two separate laboratories: one at AUA in Athens, and one at Videometer’s facilities in Herlev, Denmark. Each laboratory used its own unit of the same portable MSI instrument. This cross laboratory, cross device design was intended to test whether results remain consistent when the method is applied in different settings with different equipment of the same type, a factor that can affect reproducibility in spectral analysis.

The instrument used in this study was a prototype of what is now known as VideometerLite+. At the time of the trials, it captured images across 7 wavelengths, ranging from 405 nm to 850 nm.

The current version of VideometerLite+ has since been updated. It now captures 15 wavelengths, spanning 405 nm to 960 nm, a wider spectral range than the prototype used in this study. The instrument remains portable and straightforward to operate, capturing a full multispectral image within seconds. Each individual instrument is calibrated to ensure that measurements stay consistent.

Key Findings: Freshness, Thawing, and Origin Detection

The spectral data were analyzed using three predictive machine learning models: Partial Least Squares Regression (PLS-R), Support Vector Machine (SVM), and k-Nearest Neighbors (kNN). Each model processes data differently, so comparing them allowed the researchers to identify the best performing approach for each task. Three separate questions were examined.

• Freshness. The models were used to predict microbial load from spectral readings. Performance was high across all three models, with the 460 nm band consistently identified as the most informative wavelength for tracking spoilage.
• Fresh versus thawed. The models were also used to classify whether a sample had previously been frozen. Accuracy reached up to 95% for the Danish samples, but was lower and less consistent for the Greek samples. The authors attribute this difference to variation in thawing conditions between the two laboratories rather than a limitation of the technology.
• Origin (Greek versus Danish). Origin classification produced the strongest results, with several models reaching 100% accuracy. The 660 nm and 850 nm bands were the most useful for this task, likely reflecting differences in fat and pigment content between the two origins.

Practical Implications for Poultry Quality Control

The findings indicate that a single portable multispectral scan can support several quality and authenticity checks at once, including detection of early spoilage, identification of chicken that has been frozen and sold as fresh, and verification of origin claims. This kind of combined, non-destructive assessment is relevant to ongoing concerns around food fraud and supply chain transparency.

The authors note that this is a single study and that further validation is required. Real world conditions, such as temperature fluctuations during transport, have not yet been tested. A wider range of chicken products and packaging types, along with larger and more varied sample sets, would also be needed before broader conclusions can be drawn.