|
Illustration of the contact free probe |
A number of different techniques have been applied to this problem earlier. Near Infrared (NIR) spectroscopy has shown the most consistent and promising results among other techniques like Magnetic Resonance (MR) and impedance measurements.
In this case we have studied the range from visible light up to 1100nm. The water content can be estimated from e.g. the strength of the water absorption line at 970nm.
Split cod is not a straightforward sample to measure with optical spectroscopy. It is covered with a thin layer of strongly light scattering salt, the skin is coloured and reflective, fins and bones get in the way, and the fish meat is fibrous and strongly non-isotropic.
Reflection measurements have been tested, but results are strongly influenced by surface effects (salt, skin, and fins). In transmission measurements these problems would be smaller, but the dominating light scattering makes good quantitative measurements unachievable. Transflection measurements, i.e. where light enters the fish at one point and is collected at another, show however useful results, as surface effects are reduced and scattering contributes to a longer and deeper path within the fish meat. Transflection measurements traditionally require a contact probe, which is undesirable in industrial applications. Preferably, measurements should be automated and contact free.
Our first step was to develop a contact free probe to be used with the contact probe (figure). Despite far lower SNR the contact free probe provided as good results. The second step was to adapt the method to an industrial setting. For this purpose we reengineered a waste sorting scanner (see picture below) which allowed us to make ~10000 measurements pr sec, thus having no problem mapping the water content of a fish at 1 m/s. With the scanner we obtain as good or better results than with lab instruments (±0.8%). The scanner was demonstrated at Westfish in May 2004.
The water content measurement will be complemented with a machine vision system for automatic characterization of other quality parameters like blood or liver stains, yellowness and rupture.
The scanner has also been tested to measure e.g. fat content of salmon and meat with promising results.
This work has been done in close cooperation with Matforsk and was supported by the Norwegian Research Council (NFR) and Fishery and Aquaculture Industry Research Fund (FHF).