Introducing Live Face Matching
Introduction
Live Face Matching (LFM) from Hitachi analyzes live video to recognize registered individuals for security or operational purposes. Highly accurate and able to run on a variety of different camera feeds, LFM is a powerful tool for law enforcement, corporate security, identity-based operations, and customer services. High accuracy, frame rate capacity, and affordability make LFM a clear choice for your facial recognition solution.
Feature highlights
High matching accuracy
Many solutions use only a single face matching pattern. When matching the input video to the facial image of the target person, LFM identifies positions of each facial feature, and matches by extracting and comparing these features in the same position. Up to 6 patterns of face images can be registered per person. This improves matching accuracy in the field, where faces are seen in multiple directions and with various positions of shadows. This is a superior approach to using only a single pattern.
High speed matching
LFM can extract multiple faces from a single image in parallel and match in real-time (5fps), processed at high speed max of 60 faces/second with a single server input of 4 cameras running in parallel.
Robust solution for real-world environments
In the real world, faces shift direction and appear differently due to lighting, shadows, etc. LFM is a robust solution that is built for these environments, using existing IP video surveillance footage, or using new cameras (camera positioning will still affect accuracy, as is the case for any facial recognition solution).
Search stored footage
In addition to matching using live footage, LFM allows uploading of movie files to match faces in historical footage as well.
Auto alert reporting and configuration
LFM can report an alert to external devices, and the user has the option of switching the matching function on or off and set the threshold for facial similarity to match the monitoring environment or degree of similarity they are seeking.
Software development support
An SDK is provided to support development in the user’s application system through Web API command. LFM alerts are integrated into Hitachi Visualization Suite (HVS) and market leading Video Management Systems (VMS) for full situational awareness and operational intelligence.
Solution components
Key components
The following table lists key solution components for a Live Face Matching system in a basic setup. See the figure in Basic setup.
|
Device |
Main function |
|
Identify server PC |
Performs identification using live images from cameras, and if registered person is alerted, alert it to operation client PC. The device performs registration, deletion, etc. on to person DB according to the requests from the operation client PC. |
|
Operation client PC |
Displays the alert from the identify server on a screen. In addition, operates registration, deletion and search. The operation client is grouped into the management client and identify client by the function. For details, see Chapter 8, “Installation Procedure (Operation Client Software)”. |
|
Camera |
Takes images in the surveillance area and distributes video data to the identify server PC. |
|
SW (L2/L3 Switch) |
Connects to a camera, identify server PC, and operation client PC to form a network. |
|
Uninterruptible power supply (UPS) |
In case of power failure, shuts down the identify server PC safely to prevent damage to the database. |
|
Printer |
Prints person information registered in person DB, and alert information. |
|
CBC VMS PC |
Relays images from cameras except Hitachi ones, and allows cameras to communicate with the identify server. |
Component specifications
|
Related device name |
Specifications |
|
Identify server PC |
See xxxxxxxx |
|
Operation client PC |
See xxxxxxxx |
|
Camera (including Web encoder) |
The following cameras (Web encoders) have been verified to work normally. Hitachi Kokusai Electric cameras: KP-IP1000, KP-IP1100, KP-IP1000HD Hitachi, Limited cameras: DI-CB200, DI-CB210, DI-CD200 Hitachi Kokusai Electric Web encoders: PT-IP100T, PT-IP150T, PT-IP1900T |
|
SW (L2/L3 Switch) |
The following switch has been verified to work normally. CISCO: WS-C2960S-48TS-L |
|
Uninterruptible power supply (UPS) |
Requires a sufficient capacity to keep supplying power for about 5 minutes for the identify server to finish operations. The following UPS has been verified to work normally. GS YUASA: Acrostar THA1000R-10/El-LN |
|
CBC VMS PC (optional) |
CBC VMS “ZNS NVR” is the software CBC, Inc. provides. See the manual CBC, Inc. provides. The following version of “ZNS NVR” has been verified to work normally. ZNS NVR Product Version 2.4.2.1479 |
Basic setup
The following figure shows an example of basic connection. You may set up your system differently as long devices connection is done in accordance with the specifications of each device.
Large-scale setup
The following figure shows an example of device connections including multiple identify servers and clients. You may set up your system differently as long devices connection is done in accordance with the specifications of each device.
|
Device |
Specifications |
|
Identify server PC |
Up to 50 units without alert redundancy (see Alert notification redundancy for more information on redundant setup of alert notification) |
|
Operation client PC |
Up to 5 units Management client PC: 1 unit Identify client PC: up to 4 units |
|
Camera |
Up to 200 units (If the identify server PC is 50 units. Up to 4 cameras can be connected per one identify server.) |
Alert notification redundancy
Redundant configuration of alert notice destination is constitution that sets multiple operation client PCs as the alert notice destination of one identify server PC. (See the following figure.)
Up to 5 alert notice destinations can be set per identify server. A system can have up to 10 identify servers (with up to 40 cameras), with each configured to send alerts to up to 50 clients.
