https://www.euro-online.org/enog/inoc2007/Papers/mac-slots.html https://www.euro-online.org/enog/inoc2007/Papers/m https://www.euro-online.org/enog/inoc2007/Papers/mac-slots.html

When oranges go bad

Although there’s nothing worse than picking up a juicy orange from a fruit bowl to find it covered in blue or green mould, at least the mould provides a clear sign that the orange has gone bad. But while this mould is easy to see among a few oranges in a fruit bowl, it’s much more difficult to see among the thousands of oranges processed in packing houses.

This is exactly where you want to spot the first signs of mould, though, because lots of oranges in close proximity provides ideal conditions for the fungus Penicillium digitatum, which is responsible for the mould, to spread. The mould can be made easier to spot by taking advantage of the fact that it shines under UV light, but this still requires visual checking, while long-term exposure to UV light could potentially prove harmful to workers in the packing houses.

So a team of Spanish researchers, led by Josiva Blasco at the Valencian Institute of Agrarian Research, decided to see whether visible-NIR spectroscopy offered a better method for identifying mouldy oranges. They took 117 mandarin oranges and inoculated 67 of them with P. digitatum, and then compared the visible-NIR spectra produced by healthy rinds and mouldy rinds, finding that each type of rind produced substantially different spectra.

Next, they tested the ability of various different statistical techniques to reduce the complexity of the spectral data and then generate a model that could classify oranges as either sound or mouldy. Interestingly, they found that applying scatter-correction methods to the data reduced the accuracy of the resultant models. As these scatter-correction methods are designed to remove variations in the spectra caused by physical differences between samples, this suggests that the mould produces greater structural rather than chemical changes to the rind.

As they report in the Journal of Food Engineering, the best model was produced by factor analysis working purely with NIR spectra, with this model able to identify 100% of the healthy oranges and 94% of the mouldy oranges. Still, there is a lot work to do before NIR spectroscopy can be deployed in packing houses, not least finding a way to scan the entire surface area of lots of oranges at the same time.

Blog tags: 
Techniques: 

News Blog Archive

Latest NIR Events

7 May 2019 to 9 May 2019
Beijing, China
8 July 2019 to 12 July 2019
Auckland, New Zealand
15 September 2019 to 20 September 2019
Gold Coast, Australia