Conference paper
Using image analysis to monitor biological changes in consume fish
The quality of fish products is largely defined by the visual appearance of the products. Visual appearance includes measurable parameters such as color and texture. Fat content and distribution as well as deposition of carotenoid pigments such as astaxanthin in muscular and fat tissue are biological parameters with a huge impact on the color and texture of the fish muscle.
Consumerdriven quality demands call for rapid methods for quantification of quality parameters such as fat and astaxanthin in the industry. The spectral electromagnetic reflection properties of astaxanthin are well known and have in previous studies been shown to change as a function of astaxanthin concentration.
This may be utilized to quantify the amount of astaxanthin contained in salmonid fishes by assessing a spectral measurement of the fillet. Existing ways of assessing the amount of astaxanthin and fat in salmonid fishes is based on highly laborious chemical analysis. Trichromatic digital imaging and point-wise colorimetric or spectral measurement are also ways of estimating either the redness or the actual astaxanthin concentration of the fillet.
These methods all have drawbacks of either cumbersome testing or lack of spectral or spatial information. The use of multispectral imaging to assess the variation in fat and astaxanthin concentration both between fish and within fish is investigated. Since reference values cannot be obtained corresponding to pixel-wise measurements, sub areas of the fillet are predicted and averaged in order to establish correlation to reference measurements.
In the present experiment salmon fillets were sampled at different locations. Each sample consists of a biopsy which was cut in three layers in order to understand the depth distribution of astaxanthin as well as the spatial distribution. For each layer in each sample, chemical fat and astaxanthin determination was carried out as a reference value and a multispectral image was acquired to establish a correlation.
A prediction model has been calibrated to predict the astaxanthin concentration on a pixel-wise level resulting in an astaxanthin-map of the fillet reflecting the intra fillet variation in astaxanthin concentration. The model is validated according to spatially varying samples across the fillet.
Language: | English |
---|---|
Publisher: | Chalmers tekniska högskola |
Year: | 2011 |
Proceedings: | 41st WEFTA Meeting,Gothenburg |
Journal subtitle: | Abstracts |
Types: | Conference paper |
ORCIDs: | Frosch, Stina and Nielsen, Michael Engelbrecht |