Face-storing database software
February 01, 2009
Facial recognition processing requires a high-powered, small memory footprint real-time database.
Recognizing enterprise application needs
Selecting the combination of hardware and commercially
available real-time database software is a decision that almost every embedded
systems designer must make. Particularly when designing facial recognition
systems, engineers want database products that meet the varying needs of
embedded systems used in real-time enterprise applications.
L-1 Identity Solutionsí Bioscrypt Division teamed up with
McObject and Texas Instruments (TI) to address the challenges of creating a
facial recognition system. Using the McObject eXtremeDB Fusion embedded
database combined with a TI TMS320DM6446 digital media processor, Bioscrypt
developed a small, high-performance face reader product.
The VisionAccess 3D face reader can store up to 60,000 facial
templates and verify a personís identity in less than a second. The system
accomplishes key computing tasks including image processing and facial search
and matching on the device, eliminating its back-end industrial server.
Leveraging the embedded platform saves on shipping costs, reduces installation
complexity, and minimizes space requirements, making the security technology
unobtrusive in high-traffic workplaces such as the Venetian Macao-Resort-Hotel
in Asia, which uses the face readers to authenticate its 12,000 employees.
This project illustrates how benefits such as performance,
ease of development, small form factor, and low production costs are often
secured early in a productís life at the stage of comparing and choosing
technology building blocks. Although many devices are connected via the
Internet, having a local database that is fast and flexible enough to meet the
many configuration options found in embedded computing systems is a must.
Maintaining a secure connection with emerging enterprise solutions is critical
for future success.
Graves, cofounder and CEO