The first approach measures pixel values at key locations on a person's face, which are then compared to a database of sober and drunk face scans. Similar technology has been used at international borders to identify people infected with virus such as influenza or SARS.
The second approach compares temperature differences between parts of a person's face. For instance, a drunk person's nose will typically be hot, while the forehead tends to be cooler. This approach requires the algorithm to identify parts of the face present in the thermal image, but does not require the context of an outside database to function.
Combining the two techniques into one technology could allow the quick scanning of people whose sobriety is in question. Law enforcement use is the most obvious example, but one day bartenders might have scanners they use to decide whether someone has had enough to drink and needs to go home.
