Mahmoud Ouazine

Hitachi VMware-Based Solution  for Live Face Matching (LFM)

Blog Post created by Mahmoud Ouazine Employee on Oct 20, 2016

Real time face matching and recognition refers to the task of locating human faces in a video stream and identifying the faces by matching them against the database of known faces.

To support public safety solutions that foster safer, smarter, more efficient communities through connected intelligence, Hitachi Data Systems with a collaboration with Hitachi Kokusai has designed VMware-Based Solution for Live Face Matching . This solution  helps law enforcement and emergency management personnel prevent and collaboratively address public safety situations

The  Hitachi VMware-Based Solution  for Live Face Matching provides real-time high-speed matching of stored images. LFM matches a facial image that is extracted from the live video of surveillance cameras immediately to the facial image of the target person’s database that has been previously registered. LFM uses live multi-camera video to analyze stored video and images to match and report subjects of interest, critical infrastructure sites strengthen security and increase threat detection which are essential to ensure safety.

Hitachi VMware-Based Solution  for Live Face Matching is  cost-optimized solution that can be deployed in public and commercial buildings, hospitals, banks, railways, cruise lines, airports, and many others where continuous pro-active threat detection is essential.

  • LFM detects and reports subject facial images in real-time.
  • LFM immediately matches facial images extracted from live surveillance camera feeds with image databases at high speed (max 60 faces/sec).
  • LFM alerts and reports detailed information of matching results
  • Flexible Camera Selection (Other than Hitachi manufacturer’s IP camera can be adaptable by combining with the optional interface module)

Hitachi VMware-Based Solution  for Live Face Matching consists of the following :

  • Rack Optimized Server for Solutions, 2U Single Node with at least 128 GB RAM, 2 × 4 TB (RAID-1) and  2 × 10 GbE ports  for ESXi , LFM Server VM and ZNS Server VM
  • VMware  (Scalability and solve cost  issues)
  • Cameras
  • Network infrastructure
  • LFM Software (Facial Identification)
  • ZNS Software (Capturing , transcoding and distributing Video Stream)
  • Generic Compute System Server for LFM client (Alerts and related information)

The bellow figure shows a schematic Configuration Diagram for this solution when using Camera simulator software

The cameras are deployed in  places where continuous pro-active threat detection is essential. LFM client sequentially reports subjects registered in the image database. In case the facial similarity rate is greater than or equal to the fixed value (threshold) , LFM can report an alert to the operator (that is, security personnel monitoring the system). LFM consists of the identify server software running on a identify server with Linux, and the client operation software running on a client PC with Microsoft® Windows®. The Identify server software can identify face matches at up to 60 faces per second.  LFM also provides a movie file identify function, that can match any of the uploaded video files to the server.

LFM server as a virtual machine (VM) has tested  been tested in a VMware environment  and supported. Hitachi best practices recommend that customers can use LFM VM server on a Rack Optimized Server for Solutions, 2U Single Node hardware configured with 30 vCPU and 96 GB of memory and ZNS VM server configured with 16 vCPU and 32 GB of memory.

Having LFM server and ZNS server as virtual machines running on VMware ESXi on a Rack Optimized Server for Solutions, 2U Single Node hardware offers customers affordable LFM features with high performance, large face matching database, and an optimum frame rate process of 5 fps.

 

More information can be found in  Public Safety Scalability Testing LVR white paper at the following link:

https://www.hds.com/en-us/pdf/white-paper/hitachi-lab-validation-report-vmware-based-solution-performance-for-lfm.pdf

Outcomes