Agnar Holten Sivertsen

Automatic Nematode Detection in Cod Fillets (Gadus Morhua) by Transillumination Hyperspectral Imaging

Traditional quality control of cod fillets is currently made by manual inspection on candling tables. This is a time consuming and expensive operation, contributing to a significant share of the cost with cod fillet production. In this study, transillumination hyperspectral imaging was implemented as a method for automatic nematode detection in cod fillets moving on a conveyer belt, and evaluated on industrially processed cod fillets. An overall detection rate of 58% of all nematodes (N= 922), with detection rate of 71% and 46% for dark and pale nematodes, respectively, is reported. This is comparable, or better, than what is reported for manual inspection under industrial conditions. The false alarm rate was high, with 60% of the fillets reported with one or more false alarms. These results show that the method is promising, but needs further refinements to reduce the false alarm rate and increase the imaging speed from 25 to 400 mm/s. Practical Application: Manual inspection of cod fillets is a huge bottleneck for the industry, accounting for half the production cost with cod fillet processing and reducing the processing speed. Transillumination hyperspectral imaging has the potential to reduce the manual labor required for cod fillet inspection and hence reduce the cost and increase the end product quality.

Effects of single wavelength selection for anisakid roundworm larvae detection through multispectral imaging

The occurrence of parasites in fillets of commercially important fish species affects both food quality and safety. Presently, the detection and removal of nematode parasites is done by inspection on a light table (candling) and manual trimming of the fillets. This operation is costly and time-consuming and is not effective for detecting and removing all the nematodes in the fillets. In the last decades, several alternative methods have been proposed, but these methods have failed to replace the candling method. A newly described method called imaging spectroscopy has produced promising results because the operator can record both spectral and spatial information from an object. In this work, we studied single-wavelength bands from a spectral image. Discrimination between nematodes and other objects in the fillets is dependent on the level of contrast. Quantification of the contrast in such images revealed that the level of contrast varied when different wavelengths were selected, and these variations are correlated with the absorption properties of the nematode. Visible light scatters greatly in fish muscle, generally complicating the detection of nematodes. In this study, light scattering was used in a way that reduces the background complexity in spectral images. When light scattering properties were used in a wavelength range different from the bulk of the nematode light absorption, spectral images with significantly higher contrast were produced.