On-line transflectance NIR imaging of foods

Heterogeneity of foods is a challenge with regard to instrumental sampling. On a stream of fairly homogenous material, such as ground meat, an average reading can be obtained by accumulating instrument readings over a large part of the sample. However, in many cases we need to characterize not a stream/batch, but single products such as fish fillets, cheeses, meat trimmings etc. Since such products are often heterogeneous, it is not necessarily of interest to measure on one spot, but as much of the product as possible. One way to cope with this using spectral measurements, is to apply spectral imaging. This approach gives images with a spectrum in each pixel, and therefore a much better representation of the sample in terms of internal variation etc. Spectral images enable more precise and representative sampling.

Together with Sintef ICT and the Norwegian company Qvision AS, we have developed an industrial on-line VIS/NIR spectroscopic imaging scanner. This is a non-contact system, which can be positioned over any conveyor belt. In the NIR region, the measurements are done in so-called transflectance mode: The samples are illuminated along a line, and the transmitted (back scattered) light is measured along another line some centimetres away. The direct reflected light is suppressed, and in this way we assure a rather deep optical penetration (20-25 mm), and representative spectral sampling the products.

The system has been commercially developed to measure fat and pigment in salmon fillets, and water content in dried salted cod. The capacity of the system is about 60 fillets per sec. Since the system is imaging, not only the average chemical composition can be estimated, but also the internal distribution of the chemical components in the product. The system is presently being further developed for the meat industry, as well as many other applications.


Fig. 1: Principle of operation for online transflectance system




Fig. 2: Prediction of water contents in dried and salted cod. An image of a dried and salted cod for comparison to the right.

Fig 3: Pixel-wise prediction of fat content in a salmon fillet


Staff
Jens Petter Wold
Vegard Segtnan
Martin Høy
Bjørg Narum
Frank Lundby


Collaboration
Trond Edvardsen, Qvision AS, Oslo Norway
Jon Tschudi, Sintef ICT, Optical measurement systems and data analysis , Oslo, Norway

Financial support
Matforsks strategical research programme:
Effective assessment of food quality by rapid spectroscopic and sensor techniques
Project period: 2005-2008
Project leader: Jens Petter Wold
Financed by: Norwegian Agricultural Authority

Matforsks strategical research programme:
Biostatistics and bioinformatics
Project period: 2005-2008
Project leader: Achim Kohler
Financed by: Norwegian Agricultural Authority


Publications
Wold JP, Johansen IR, Haugholt KH, Tschudi J, Thielemann JT, Segtnan VH, Narum B, and Wold E. 2006. Non-contact transflectance near infrared imaging for representative on-line sampling of dried salted coalfish (bacalao). J Near Infrared Spectrosc. 14: 59-66.
Wold, J. P., Jakobsen, T., and Krane, L. 1996. Atlantic salmon average fat content estimated by near-infrared transmittance spectroscopy. J. Food Sci. 61(1): 74-77.
Wold, J. P. and Isaksson, T. 1997. Non-destructive determination of fat and moisture in whole Atlantic salmon by near-infrared diffuse spectroscopy. J. Food Sci. 62(4): 734-736.


Links
Effective assessment of food quality by rapid spectroscopic and sensor techniques
Biostatistics and bioinformatics
Qvision
Sintef ICT, Optical measurement systems and data analysis

Back      Print page    Home