In partnership with DIAM Bouchage, CEA List, an institute of CEA Tech, has developed technology based on artificial intelligence analysis of X-ray images to detect defective corks in a timely and accurate manner.
Although cork is often considered ideal for long-term wine preservation, sometimes a wine turns corked or oxidized when opening a bottle... To properly fulfill its function, a cork must allow just enough oxygen to pass through to both allow the wine to ripen and guarantee its taste quality over time. However, it is impossible to effectively appreciate the sealing of a cap with the naked eye.
Diam Bouchage has enlisted the help of the Carnot CEA List Institute to develop an automated and reliable method of sorting corks from the casing of cork oak bark. The selection is made on images produced by X-ray tomography that allow to visualize the cap in its volume. To analyze these images, the researchers set up a machine learning algorithm adapted to classification issues. By comparing the characteristics obtained by imaging (number and distribution of lenticelles, growth lines, cork density,...) with data on the rate of oxygen transfer carried out over the long term, they have set up a model that allows to automatically assess the sealing of the cork, guaranteeing the quality of the cap.
The researchers were able to significantly improve the classification of plugs: 75% of them classified correctly in just a few seconds. Requiring only two frames per cap, this method can be considered to deploy automated online non-destructive control of natural cork caps. Work continues to achieve 100% good classification, a mandatory criterion for the ceding of great wines.
Visit the site: