No more dodgy prawns

There’s nothing quite like a tasty, succulent prawn. Occasionally, though, you’ll come across a prawn that’s a bit tough and chewy, especially if it's been frozen at some point. This disappointing eating experience could soon become a thing of the past, however, because Chinese food scientists have shown that NIR hyperspectral imaging can identify such dodgy prawns before they go anywhere near your mouth.

The scientists from South China University of Technology, Guangzhou, led by Da-Wen Sun (now at University College Dublin in Ireland), analysed 100 peeled prawns with hyperspectral imaging, a form of spectroscopy that produces images in which every pixel contains spectral data. They also measured three physical properties of the prawns, their hardness, gumminess and chewiness, with a mechanical testing device that compressed the prawns. Up to now, such mechanical devices, along with manual inspection by trained assessors, have been the only way to measure the physical properties of prawns without eating them, but both approaches are very time consuming and can also damage the prawns.

Sun and his colleagues thought that NIR spectroscopy might make an effective, and much faster, alternative, because these physical properties are generated by underlying chemical properties, such as protein and collagen composition, that can be measured by NIR spectroscopy. Furthermore, hyperspectral imaging should allow them to monitor how these physical properties vary over the surface of prawns.

As a first step, the scientists used a mathematical technique known as the successive projections algorithm to select those NIR wavelengths that were most responsive to changes in the physical properties of the prawns. Next, using the non-linear statistical technique known as least-squares support vector machines, they produced a predictive model relating spectral data at these wavelengths to the measured physical properties.

As they report in the Journal of Food Engineering, this model turned out to be fairly accurate at predicting all three physical properties from the spectral data for individual prawns, with accuracies of 81−86%. As hoped, Sun and his colleagues were also able to map these properties over the surface of the prawns, producing coloured images in which red denoted high values for the physical properties and blue denoted low values. Using these images, they were able to show that the heads of peeled prawns are generally harder than their tails.

Blog tags: