This website uses cookies. By continuing to browse the site, you confirm your consent to the use of these files.

+7 (499) 238-01-32
Mon–Thu: 09:00–18:00

Fri: 09:00–17:00
Contact
Address 119180, Moscow, 1st Khvostov Lane, 11A
Working hours Mon–Thu: 09:00–18:00, Fri: 09:00–17:00
ASRL Biovizum — face recognition system

Automated face recognition system «Biovizum»

The automated face recognition system «Biovizum» provides the possibility of contactless identification of a person's face.

Contents
  1. General description of the ASRL Biovizum
  2. Functions of the system
  3. Face detection
  4. Face recognition
  5. Verification and identification
  6. Related products and services

ASRL «Biovizum» is an automated face recognition system using a convolutional neural network with more than 20 layers for high-precision identification of a person. The system is designed for contactless identification of a person by face in real time.

The face recognition module is designed for automatic identification of a person by photo image and can be used both for verification (one-to-one comparison) and for identification (search in a face database).

Functions of the ASRL «Biovizum»

  1. detection of a face (or several faces) in a photo image or video stream;
  2. face recognition (extraction of a feature vector from a biometric sample of a face image);
  3. determination of the degree of similarity (verification) of two samples;
  4. search (identification) in the database of registered samples.

The ASRL «Biovizum» uses its own mathematical face recognition algorithm based on a convolutional neural network with more than 20 layers of neurons.

Face detection

The face detection function determines all faces present in the image (frame) and performs an assessment of the face image quality, which can be used to select the best image for further processing. The function additionally computes the position of facial landmarks, which allows monitoring and taking into account a person's facial expressions, as well as determining the shooting angle.

Face recognition

As a result of face recognition, a template is constructed, which is a set of numbers (vector) that determine the characteristic features of a person's face. The feature extraction process is stable over a wide range of head rotation and tilt angles, as well as lighting conditions, since the images are preliminarily brought to a conditionally frontal type using the data obtained during detection.

As a result of feature extraction, the person's ethnicity, age and gender are determined. The presence of facial hair (beard, mustache), makeup, glasses, etc. is taken into account.

Verification and identification

Verification and identification are based on the process of comparing templates with each other, which consists in calculating the «distance» between feature vectors in a multidimensional space. A low «distance» value means a high degree of similarity of the original face images.

To conduct a search (identification) against large face databases containing millions of photographs, optimization methods of preliminary selection of match candidates are used. As a result, the search can achieve millions of comparisons per second.

The module has the following capabilities:

  • detection of a face (or several faces) in a photo image;
  • construction of a face signature (a unique combination of numbers describing the face) from a photo image;
  • signature comparison (based on two signatures, a decision is made whether they belong to the same person or not);
  • search by photo image in the face database;
  • assessment of face image quality (check for glasses, beard, masks, assessment of lighting and angle);
  • integration with video surveillance systems for face analysis in real time.

ASRL «Biovizum» is designed for use in access control tasks and monitoring of people's movement and behavior.

Related products and services

Related technologies