Russian Feedback Home
S O V A

SOVA




is designed to form and store a database of those wanted by law enforcement agencies and for the identification of citizens...


IDENTIFICATION

Identification






system for law enforcement authorities intended for prompt identification of a person made by way of remote access to a central server from any type of officers work places.


CHECKPOINT

CHECKPOINT




is a automated information storage and retrieval system for detection of persons using forged passports or visas.


PHOTOSYSTEM







Facial identification of citizens applying for national ID’s and passports to reveal individuals wanted by the authorities.


CONSUL

CONSUL




system implements a set of safety measures related to the identification of criminals and terrorists.


MIGRANT

Migrant




is intended for the automisation of all records, digital photographs and the creation of a database on the electronic carriers of internationally wanted persons.


Videostream

BIOMETRIC SURVEILLANCE SYSTEM

DESCRIPTION


The VIDEOSTREAM system is intended for video surveillance in crowded places. Its function is to retrieve each facial images caught by the video camera from a biometric database. With a high degree of accuracy, the VIDEOSTREAM system recognises the wanted person when he/she crosses the area covered by the monitoring system which can be installed in transportation hubs, subways, as well as in streets, squares and stadiums. The system will automatically give notification of the identified person's location to authorized law enforcement officers.

The VIDEOSTREAM system is a distributed video surveillance system which incorporates the latest advances in intelligent video information processing, biometric facial identification, specific population database searches, and video archive information gathering. Facial image databases generated by the SOVA system can used by the VIDEOSTREAM system as retrieval databases.

PRINCIPAL OF OPERATION

The VIDEOSTREAM system is a high-end distributed video surveillance system consisting of facilities for video capture using stationary and scanning day and night vision cameras, stationary video cameras for the recognition system, video digitalization, delivery and processing systems, recognition servers, as well as monitoring, administration and control of the distributed systems.

The VIDEOSTREAM system is principally designed for automated facial identification in crowded places such as subways, stadiums, squares, etc. The system processes the video flow of incoming images (24 images per second) and automatically performs real-time selection of those facial images which are most suitable for verification out of the video flow. The selected facial images are transmitted to recognition servers (they may be either local and remote), and the recognition servers perform instant (fractions of a second) comparison of these images with facial images of wanted terrorists and criminals. If the facial image taken by the video camera among a crowd of people resembles a wanted person, the VIDEOSTREAM system informs the authorized law enforcement agency which will make a decision on any further required action. The estimated response time is no more than 10 seconds. Thus, the system solves the following problems: recognition of faces in a flow of people, search for persons in video archives and search for persons through a database containing photos and personal data.

The VIDEOSTREAM system enables automatic identification among a flow of people, verifying the obtained images against wanted persons databases. The high accuracy of identification and the high speed of system operation makes the effectiveness of using the biometric video observation system in combating crime and terrorism a real possibility.

When configuring a video surveillance system, a number of conditions should be considered to achieve error-free identification of criminals in a crowd. Recognition system stationary video cameras should be installed in the places of traffic where facial images of best quality can be taken for example, at stadium turnstiles, concert hall entrances, subway escalator step-on/off places, etc. The video capture area should be uniformly illuminated without shadows. If necessary, the illumination may be on a 24-hour basis. For example, using a backlit poster produces quite a good effect. In the place of installation, some elements may be used to attract the attention of people: the same bright poster can cause turning/raising of the head at the angle you need. The method of mounting each specific video camera and the necessity for using a ⌠set of hidden settings■ should be determined at each location.

In the regional operative centre of the VIDEOSTREAM distributed video surveillance system, information about the wanted person's location in a certain place and general video information from panoramic video cameras can collectively allow law enforcement officers to make an appropriate decision on detaining criminals detected by the video observation system.

TECHNICAL PARAMETERS

number of cameras concurrently operating on one server 32 pc
number of simultaneously caught and processed facial images 3 - 5 persons (frame format 640x480 pixels)
video set processing speed 24 frames per second

    ∙  retrieval rate depending on database size:

Database size Time of search in seconds
10,000 images 0.06
100,000 images 0.14
1,000,000 images 0.73
2,500,000 images 1.49

    ∙  probability of identification (Provided the obtained facial image is of good quality)

Diagram 1.
Automatic identification mode (the system decides on identification results).

False Accept
Rate(FAR), %
0 1,0E-08 1,0E-07 1,0E-06 1,0E-05 1,0E-04 1,0E-03 1,0E-02 1,0E-01
Correct Identification
Rate (CIR), %
23.8 23.8 30.2 36.1 46.3 63.3 80.5 94.4 99.8

Diagram 2.
Expert identification mode (it is the expert's prerogative to make a decision on identification results).

Database's size Position
1 1-5 1-10 1-20 1-50
Probability of identification
10,000 images 96.5 99.3 99.7 99.8 99.9
100,000 images 91.3 96.2 97.5 98.5 99.2
1,000,000 images 84.7 91.2 93.3 94.7 96.3
10,000,000 images 81.6 88.2 90.5 93.1 94.8

RECOGNITION SYSTEM VIDEO CAMERA REQUIREMENTS

- Resolution: over 450 lines;
- Objective lens with optical magnification: no less than x22;
- Sensitivity: no less than 0,03-0,5 lux, lens aperture F1.2, signal-to-noise merit 46 dB for black-and-white    cameras.

Since there are many camera manufacturers, a wide range of video input devices can be used, provided the requirements of the technical parameters are met.
Biometrics Exhibition and Conference 2008

Copyright © 2000 - 2010 Asia Software Ltd. All Rights Reserved.