Facial recognition technologies affect us all as we clearly each have the one thing required for these powerful biometric tools to identify us, a face.
With facial recognition becoming more accurate, we see it more in our lives, as many use it to access their phones with a mere glance.
In addition, government agencies are now effectively employing facial recognition technology to search for and identify criminals.
As we integrate facial recognition into our daily lives, understanding this technology is important nt. In this article, we will discuss the following facets of facial recognition:
- Current issues with facial recognition
- How facial recognition works
- How we evaluate the effectiveness of facial recognition systems
Come join us
Are you interested in working with these types of innovative technologies? Join our team!
Facial Recognition Issues
The primary issue of facial recognition regards our privacy. As many of us now have pictures of ourselves on our social media profiles, we become increasingly vulnerable to private companies or government agencies using this technology to identify us without our consent. As a result, facial recognition has been the subject of many privacy rights cases and claims of civil liberties violations.
Facial recognition also has its fair share of technological flaws such as difficulty accounting for changes in our appearance. For example, a facial recognition system may fail in making a positive match between photos of someone before and after they have undergone significant weight loss.
Other changes in appearance such as growing or removing facial hair, aging, and the presence or absence of accessories on the face (e.g. hats, sunglasses, contact lenses instead of eyeglasses) can also trick facial recognition systems.
Despite these issues, facial recognition remains reliable enough for many applications.
For instance, the e-Passport infrastructures of many nations use facial recognition to identify travelers. Major international airports also use large-scale facial recognition to search for suspects on terrorist watch lists.
How Facial Recognition Works
Facial recognition technology relies on the unique, genetically-determined physical features of the face to accurately identify a person.
Current facial recognition systems focus on the parts of the face which are not as prone to dramatic, unpredictable changes as described earlier (weight loss, aging, hair growth/loss, etc). These regions of the face include:
- The ridges between the eyebrows
- The cheekbones
- The edges of the mouth
- The distances between the eyes
- The width of the nose
- The contour and profile of the jawline
- The chin
Facial recognition can be applied as a fully automated system or as a semi-automated system. A fully automated system requires no human interaction to make the final verification or identification decision. Conversely, in semi-automated systems, user input confirms the computer’s decisions to ensure accuracy.
Given the potential for misidentification with this technology, semi-automated systems prove more dependable in their final decisions.
Facial recognition Process
The first step in the facial recognition process involves collecting the raw images of the individual. Facial recognition systems require the user to stand before a camera, or, if used for covert surveillance or security, they may secretly photograph a person’s face using a CCTV camera.
Next, the system locates a face within a set frame. Facial recognition algorithms detect faces with various indicators, such as skin color, the presence of a face- or head-like shape, or the detection of a set of eyes.
The computer, however, may face several challenges when locating a face in a frame such as:
- Differentiating the skin color from the background
- Detecting a face positioned at odd angles to the camera
- Identifying one face when multiple people appear in the image
Once the system collects the raw images, the software either aligns or normalizes the data to refine the images on a granular level. The refinement techniques include resizing and adjusting the face so that the software can extract the most identifiable and unique features.
Later, with mathematical algorithms, the system converts these refined images into the appropriate verification and enrollment templates. Facial recognition encounters the most obstacles in the raw image collection phase.
Facial recognition obstacles
1. Difficulty distinguishing between faces with only subtle differences.
2. Misconstruing obstructive variables in the external environment as facial features.
3. Differentiating between pictures of the same person with different facial expressions and poses.
4. Capturing a landmark, orienting features such as the eyes.
Three-dimensional Face Recognition
To compensate for these obstacles, researchers and developers have begun implementing a process known as 3D face recognition via the use of 3D imaging (three-dimensional imaging).
3D imaging
Three-dimensional imaging involves forming a shape with identifiable features from an existing 2D image. The computer applies this generated shape as a model to a 3D surface to compensate for any issues met by 2D facial recognition systems.
3D imaging technology, however, remains in the research and development phase. So, as of now, we only use it to complement 2D systems in applications requiring more precise imaging, or in environments where capturing raw images is very difficult.
Effectiveness of A Facial Recognition System
According to the International Committee for Information Technology Standards (INCITS), certain raw image requirements guarantee the effectiveness and reliability of a facial recognition system.
These requirements are as follows:
- The raw image must include an entire composite of the head, neck, and shoulders, and the individual must possess a full head of hair.
- The roll, pitch, and yaw of the raw images collected must possess a variance of at least +/- 5 degrees of rotation.
- The raw image must use only plain, diffused lighting.
- Raw images must not contain shadows.
Influencing factors
We can ensure the effectiveness of a 3D facial recognition system by meeting the following criteria:
- If using stereo imaging, the system must utilize at least two cameras, each mounted at a fixed distance.
- If using structured lighting, the facial recognition system must flash a defined, structured pattern onto the face to help compute depth.
3D systems can also use both stereo imaging and structured lighting in conjunction.
Some 3D imaging systems employ laser scanners instead of cameras to sense facial features. These scanners are very robust and effective. However, their implementation is costly. Moreover, these scanners require 30 seconds or more to capture and process the raw image of a face.
Facial Recognition Overview
As we discover more ways to implement facial recognition technology, it's important for us to understand how it works, how we can use it, and how governments and companies implement it. The use of facial recognition in our personal devices will make our lives more convenient, but the use of these systems by certain agencies raises many ethical concerns.
Facial recognition works by capturing raw images, identifying a face, and matching the facial features with a picture in a database. Inaccuracies may occur if the individual has undergone dramatic changes in appearance, or if certain elements of the photo confuse the identification algorithm. By meeting certain effectiveness requirements, we can avoid many of these problems.
3D imaging technology also compensates for many obstacles affecting a traditional 2D-imaging system’s ability to identify a face. However, this technology still requires additional research and development before we can reliably deploy it.
Stay tuned for a future article in which we will examine the mathematical models that are used in facial recognition, as well as their advantages and disadvantages.
