For each detected face in an image, face recognition attempts to find its closest match from one or more pre-trained databases. Media Server then returns a configurable number of the most closely matching faces from the databases, along with a confidence rating for each match. The results are ordered by confidence rating. Media Server's face detection module detects the faces in the image.
Before face recognition can occur, you must populate the face recognition databases with training images for the faces that you want to recognize. For each individual that you want to recognize, you must provide one or more good quality images of their face, along with an identifier for the person. You can create any number of face databases, each of which can contain any number of faces. You can use IDOL Admin to perform the training process.
Face recognition has a wide variety of uses, including:
Law enforcement – automatic identification of known suspects
Border protection / immigration control – automated entry gates at airports
Security / access control – CCTV-based access control, allowing specific people into specific zones
Personalization – PCs, home entertainment systems, and so on, can provide personalized content based on user
The accuracy of face recognition depends largely on the images that it processes. See Factors Affecting Face Recognition Performance. You can also configure face recognition for best performance.
|