Face recognition is now a very hot topic, and the market prospects are promising. With the popularization and application of biometric technology, face recognition has also unveiled its mystery. Today, the editor will take you to understand the process of face recognition technology.

Speaking of the face recognition process, in fact, in general, it mainly includes face image acquisition and detection, face image preprocessing, face image feature extraction, matching recognition, etc. Let us briefly understand each part.

Face image collection and detection

As the name implies, it is a face entry system, which is generally batch entry, and users actively upload and import their own faces through the entry entry. The other is the on-site video face collection, which is similar to applying for an ID card. The camera collects faces on the spot. When the face recognition collection needs to be within the specified camera range, the collection device will automatically search for the customer's face image.

Alipay’s face recognition payment adds live detection, requiring customers to perform a series of actions, such as opening mouth, raising head, bowing, turning left, turning right, blinking, etc. Face recognition technology is also continuously improved, and activity detection can effectively resist common attacks such as photos, face changes, masks, occlusions, and screen remakes.

Face image preprocessing is a series of complex processing such as light processing, cutting, rotation, noise reduction, filtering, enlargement or reduction of the face image collected by the system. Through these processes, the face image is composed of light or angle. , Distance, size, etc., to meet the standard requirements for facial image feature extraction, and to eliminate the influence of factors such as illumination and angle as much as possible to prepare for facial image feature extraction.

Matching and recognition

● 1:1 comparison of face recognition

 The face recognition system uses face recognition algorithms to compare the first two images, returns the comparison results according to the recognition rates of different channels, and compares them according to the set rules The image is stored in the database. Recognizable rate, intelligently solve the shortcomings of low pixels (such as chip images), backlight, side light, dimness, glasses, and a certain angle of the face.

● Face recognition 1:N contrast

The face recognition system recognizes the identity of the customer in the customer feature database through the customer’s image, and returns the customer’s relevant information, such as customer information number, name, etc. The system has a 1:N face recognition function and provides a 1:N comparison interface. It can extract feature values based on the photos sent by each system, compare them with the templates in the library, and return N people with similarity (the number of returned people can be customized).

Face recognition is currently mainly used in bank VIP customer recognition, candidate recognition in the education field, customs clearance, access control and attendance, intelligent video surveillance, etc.