How does facial recognition software work?

Any futurists among you, sci-fi fans, or those with a curious nature may have been exposed to fictional impressions of facial recognition software. Sounds like technology that’s far, far away in the future, right? Well, believe it or not facial recognition software has been commonplace for almost a decade.

In 2001, the Tampa Police Department installed police cameras, equipped with facial recognition technology in the Ybor City nightlife district in an attempt to combat the growing crime rate. Sadly, after a two year experiment the software was deemed ineffective, and the project scrapped, yet with perseverance this technology has become prominent in today’s society.

Based in Minnesota, Identix is one of many companies that are presently developing facial recognition software for mainstream use. The Facelt software has the ability to pick out an individual face from a crowd and run a comparison search with a database of stored images.

So, how exactly does this software work?

Every individual face has its own distinguishing marks and features. From peaks and valleys that make up the contours of our face (or nodal points) to distinguishing features, each face is different. Each human face has approximately eighty nodal points. The software measures these through the following: the distance between the eyes, the width of the nose, the depth of the eye sockets, the shape of the cheekbones, and the length of the jaw line. These nodal points are measured creating a numerical code, called a face print. The image is then uploaded onto a database.

Facial recognition software of the past has relied on a 2D image for comparison. In order to be effective the image needs to be ideally looking directly into the camera with little variance in light or facial expression, obviously the margin for error was very slight. Even the smallest change would result in a nullified image.

3D Facial Recognition Software

As facial recognition software has evolved 3D images have become prevalent. Capturing a real-time image, with a person’s distinctive facial features like a rigid tissue and bone structure, precise eye socket curvature, and chin prominence has become possible. These features do not change over time, and are unique to the individual.

What is the exact science of 3D Facial Recognition Software work?

Firstly, the image is detected by digitally scanning an existing 2D photograph or by using a video image to provide a live picture of the subject.

Secondly, once the software has detected a face, the system determines the position of the head and the measurements of the features. The 3D software has the potential to recognise a face up to ninety degrees, whereas 2D software requires the face to be at least thirty five degrees from the camera. The software will then create measure the curves of the face on a sub-millimetre scale and create a template.

Thirdly, once the template has been established the software will then translate the template into a unique code. The code gives the template a set of numbers to represent the facial features. Once this process is determined the image is then scanned into a database containing a wealth of 3D images to determine a match. The challenge comes from the fact that some databases still only recognise 2D images.

The final step is to verify the image. The image should be solely matched to one image in the database. The match that’s identified in the database is called ‘Image 1.1’.

This software is especially prevalent and useful for law enforcement agencies, as well as a number of other sectors. Facial recognition software is here, and proving to be a valuable asset for society.