The AI-powered “electronic nose” can smell rotten meat



[ad_1]

Scientists at Nanyang Technological University in Singapore have developed an “electronic nose” powered by artificial intelligence that accurately assesses the freshness of meat.

The system uses a barcode inserted in the food package that changes color when it detects gases emitted by decomposing meat. A smartphone app then scans the barcode template to measure the freshness of the meat within 30 seconds.

In tests on commercially packaged chicken, beef and fish samples that were allowed to age, the system predicted the freshness of the meats with 98.5% accuracy.

Co-lead author Professor Chen Xiaodong said the app could help consumers decide whether meat is fit for consumption better than a “best before” label:

These barcodes help consumers save money by ensuring they don’t discard products that are still fit for consumption, which also helps the environment. The biodegradable and non-toxic nature of the barcodes also means that they could be safely applied to all parts of the food supply chain to ensure food freshness.

[Read: Meet the AI-generated bipartisan president of the US]

Researchers say their system mimics the mammalian nose method of detecting rotting meat by detecting its gases.

In the e-nose, each barcode bar contains a dye that changes color in response to different types and concentrations of gases. These reactions create a color scheme that provides an “olfactory imprint” for the state of the meat as it deteriorates.

Deep convolutional neural networks trained on barcode images then analyze the patterns to predict the freshness of the meat.

In tests, the algorithm achieved 100% accuracy in detecting spoiled meat and 96-99% accuracy in identifying fresh and less fresh meat.

The researchers have now filed a patent for their method and are working with a Singapore farm to apply it to other types of food.

You can read a research paper about the system in the magazine Advanced material.

Posted on November 10, 2020 at 2:25 pm UTC



[ad_2]
Source link