October 16, 2014 -
Aurora announced it has released version 14.10.09 (codename Inara) of its infrared passport verification software, which is able to reduce error rates for passport identity authentication by over 25%.
Aurora’s infrared to color facial recognition software works by comparing one color photograph to another, such as a passport image and a live subject photo.
Using its unique infrared to color comparison process, Aurora is able to resolve the issue of ambient lighting conditions decreasing recognition performance.
It is also able to overcome all the other challenges that other biometric facial recognition providers face, including physical changes in the subject brought on by aging, head angle, facial expression and hairstyles.
By using infrared as the light source for live capture, Aurora’s software is able to successfully overcome the issue of lighting and therefore eliminating the need for screening, booths or specially designed areas to capture the image.
As a result, users are able to decrease the overall cost of deployment, as well as ensures that if their plans change, Aurora’s hardware can be moved to any location without affecting functionality.
Using real-world image data courtesy of Atkins’ passenger authentication scanning system, Aurora has been able to collect tens of thousands of images over the past few years, along with optimizing image comparison with real passport photos up to seven years old.
As a result, Aurora’s engine has been able to “learn” how a face can change over time and to factor in said changes to generate a consistently reliable recognition rate.
Aurora’s face recognition technology can now recognize over 94% of all passengers, despite the age of the passport, change in appearance and variations in lighting conditions.
The company points out that it has tested its face recognition technology on real-world data and not an academic dataset.
Aurora currently has other developments set for release later this year, which it says will boost recognition rates even further.