Using cameras to check your alcohol level and know if you are fit to drive? This is the project developed by researchers at Edith Cowan University, in Australia. These scientists have in fact developed an algorithm analyzing video images of the driver’s face, reports New Atlas relayed by Corey.
60 volunteers participated in this project. They were divided into three groups. Some had not drunk, others were slightly drunk and the last had far exceeded the authorized limits. For their safety, these guinea pigs were placed in a driving simulator, after which an RGB camera recorded images of their faces during the entire drive.
75% reliable
Using a machine learning-based algorithm, the researchers were able to determine which people had been drinking. To do this, the AI took into account facial movements, the direction of gaze and even the position of the head.
Read also : Lotto. She poses all smiles with her husband who won the jackpot… and files for divorce 10 days later
The scientists went further by trying to determine the participants’ alcohol levels. The tool succeeded in 75% of cases with an accuracy of 0.5 grams per liter of blood, which is the limit recommended by the World Health Organization (WHO) and imposed by the Highway Code in France.
An application ?
According to the researchers, this innovation could be integrated into an application. “The next step in our research is to define the image resolution needed to use this algorithm, explains Dr Syed Zulqarnain Gilani, lead author of the study. If low-resolution video is sufficient, this technology could be used by roadside surveillance cameras, and law enforcement could use it to prevent alcohol consumption.”
The main flaw in this tool currently lies in the fact that it analyzes the driver’s behavior when he is already behind the wheel. At a time when the intoxicated subject already poses a risk to other road users. However, the researchers believe that they can still improve the device to minimize these risks and, why not, save lives.
Learn more…
