Facial recognition technology – What is it & what is it good for?
Face recognition technologies are capable of recognizing faces in pictures or videos. When taking a selfie with your smartphone, most people have probably noticed that the photo app automatically focuses the image or the focus. It does this by recognizing and localizing the face.
However, if faces are further processed for verification methods after localization, we speak of biometric face recognition. Widespread use for biometric face recognition is to unlock one’s smartphone (Face ID). In biometric facial recognition technology, 2D, as well as 3D images, can be used. But where is the difference?
Face recognition technology in 2D
Two-dimensional (2D) face recognition techniques use measurements and calculations of specific geometric characteristics of a localized face. These may include the mouth, nose, or other characteristics. In this variant, the distances, sizes, or general positions of the characteristics are measured. For the complex computation of a template, the wavelet analysis, e.g. utilizing of Gabor transformation, can be used. For template matching (the comparison of whether person XY is indeed person XY), one or more calculated frontal images of the person must already be available in the database.
Face recognition technology in 3D
In contrast to the two-dimensional images, 3D images are created using height maps of the person in question, among other things, and are therefore calculated differently. This is made possible by a different form of “scanning”. 3D images can be captured with special cameras. Often, infrared cameras are used for this purpose, as in Xbox Kinect, which additionally allows the depth of a room to be determined and compared.
Through various methods, one of which is photogrammetry (also photogrammetry or image measurement), it is also possible to calculate a 3D image from several images or a video without an infrared camera.
Calculations that rely on 2D data can run into difficulties as soon as, for example, a shadow falls on the captured person or part of the face is obscured or “misaligned” by other causes. 3D face recognition can detect errors and work around them using already known templates. The actual face can then be “created” as a whole and is only sent to template matching afterward.
Each individual template of the database requires more memory than a template from the 2D variant. Furthermore, mathematical calculations become more complex. For this reason, the search by the system, for the appropriate match in the database takes longer.
The optimal process of face recognition – example
To illustrate the process of face recognition easily, we have come up with “Anni” and “Tom” for you. Tom and Anni are a couple. Tom is planning a big surprise for Anni. Tom wants to keep everything organizational secret from Anni. Therefore, Tom wants to set up facial recognition (Face ID) on his smartphone to unlock a folder.
Tom has successfully completed the settings. He starts a trial run. The process is repeated, only with the consequence that a template matching starts. Template matching compares the mathematical complex calculations of the existing templates. Face-ID grants Tom access to his folder. Each time Tom uses Face-ID and gets recognized, either an additional template is created or the existing one gets improved. That is why Face-ID works even if the beard has grown a bit longer again.
Meanwhile, Anni has become nosy and wants to get more information. For this purpose, Anni goes to Tom’s smartphone. Anni wants to unlock the folder and sees that it is secured with Face ID. Anni tries it out anyway. The process restarts but fails to unlock the folder. Template matching reveals thanks to good face recognition technology: Anni is not Tom. Thus, she isn’t granted access, the process gets aborted and Anni can continue to look forward to being surprised by her partner. Even a beard would not have helped Anni at this point, since Face ID recognizes identification features that are not visible to the human eye.
Examples of the use of facial recognition in companies
Facial recognition technology is already being used in many industries, and for a wide spectrum of purposes. The following is a small number of use cases of facial recognition in enterprises:
- Unlocking devices
- Granting access to rooms
- Prove presence
- Prove identity
- Perform verification
- Release transactions
- Provide approvals
As you can see, there are some possible use cases, and this is just a selection. Each company should check for themselves whether it can be used sensibly. When using facial recognition software, it is important to respect the rights of scanned persons. The GDPR also provides rules for this.
Does PayPodo use facial recognition?
At PayPodo, all private and business customers are verified (keyword: Know Your Customer). When you create an account with PayPodo, you will go through an automated verification process. During the verification process, AI-powered facial recognition technology is used. Three photos are required for the verification process:
- A photo of the front side of an approved identification document.
- A photo of the back of the ID document used by number 1.
- A selfie of yourself taken with the PayPodo app.
The artificial intelligence of facial recognition technology creates at least two templates based on the information then available. Subsequently, these will move to template matching for verification. In addition to matching the images, other information is also captured and verified (address, name, etc.). Also, human checks of the process take place regularly. PayPodo works in a DSGVO-compliant manner when it comes to verification. This means that you do not have to worry about the security of your data on the part of PayPodo.