These characteristic features include various reference points and lines, such as the distance between the eyes, the distance and angulation between the nose and the eyes, the upper edges of the eye sockets, the areas around the cheekbones and the side parts of the mouth, etc. During authentication, a digital photo is taken of the person in question, which the computer compares with the specific characteristics of the reference image and makes its decision based on that.
When matching the captured image with the reference image, a grid is superimposed over the image to locate the salient points and compare their distance, location and position. Face recognition has a moderate recognition probability that varies greatly with lighting conditions and the number of people. Its false acceptance rate( FAR) ranges from 0.1 to 2%, and its false rejection rate( FRR) ranges from 0.5% to 3%.
Facial recognition is an extremely complex application whose software uses artificial intelligence and Deep Learning. When an individual faces a camera and the image is matched with the reference image, the recognition rate is 99.9%. It becomes more complicated when people moving within a crowd are to be recognized. The positive results are reduced if the persons are illuminated by unfavorable lighting conditions or if only parts of the face can be used for evaluation due to headgear or scarves.
Face recognition is not only used for people control, but also in image processing programs where photos are assigned to persons.