When failure to match is not an option Innovative new technology for accurate matching of low-quality fingerprints India Partners • Anaxee Technologies Pvt. Ltd. • Mobile: +91-96300 20005 • Email: [email protected]• Web: www.anaxeetech.com • Contact Person: Govind
27
Embed
When failure to match is not an option Innovative new technology for accurate matching of low-quality fingerprints India Partners Anaxee Technologies Pvt.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
When failure to match is not an option
When failure to match is not an optionInnovative new technology for accurate matching of low-quality fingerprints
• NFIQ tests before and after show the improvement produced by the technology.
• Independent commercial evaluations – performance gap widens as proportion of low quality fingerprints increases.
Error % Comparison by NFIQ Classification
Confidential
Evaluation and Benchmarks-Latent
• NIST ELFT-EFS public challenge
• Latent product development started in Apr 2009
• Submitted to NIST in July 2009
• Latest Version features significant improvements
Confidential
Technology
• Innovative technology uses a new model based statistical approach to overcome many of these problems.
• New approach features;• Intrinsic mathematical representation of the
ridge flow structure.• Produces “idealized” fingerprint images that
preserve distinctive features
Confidential
Technology
• Innovative technology uses a new model based statistical approach to overcome many of these problems.
• New approach features;• Intrinsic mathematical representation of the
ridge flow structure.• Produces “idealized” fingerprint images that
preserve distinctive features • Specific model for semi-permanent features
e.g. scars or scratches
Confidential
Technology
• Innovative technology uses a new model based statistical approach to overcome many of these problems.
• New approach features;• Intrinsic mathematical representation of the
ridge flow structure.• Produces “idealized” fingerprint images that
preserve distinctive features • Specific model for semi-permanent features
e.g. scars or scratches• Capable of modeling partial prints
Confidential
Technology
Confidential
Other Examples
Confidential
Other Examples
Confidential
Other Examples
Confidential
Other Examples
Confidential
Real life deployments
• MoD contracts – Latent/AFIS SDK• Mission Critical Humanitarian projects – AFIS/Latent SDK• 30+ construction sites for work force management – Livescan SDK• Passport control project – Enrollment/AFIS SDK• Auto capture for a major hardware manufacturer – Enrollment SDK• Web-based payment authorisation – Livescan SDK• Mobile AFIS screening project – Latent SDK, Livescan SDK
To Name a few:
• 1. Major European Passport control 2. Mission Critical humanitarian project in Affica3. Ministry of Defence in the UK4. Mobile AFIS in CHINA
• Accuracy, see the datasheet, also refer to NIST PFT tests http://biometrics.nist.gov/cs_links/PFT/tables2f_040510.pdf (Letter: 2E)
• Speed: 1:1 matching 4,500 per second per processor core (Intel >2.6GHz)• Template size: between 1-2KB• Memory for the program: >50MB
Technical support during integration of SDK
• Email, phone support from the UK headoffice• Field support provided by Anaxee Technologies
Confidential
Methods for license activation on server (during testing & benchmarking).
• Licence key and online or offline activation
Licensing Model
• Per node lock licence or enrolled users
Confidential
Interoperability with various sensors (optical/capacitive/multispectral etc.).
• Support all sensors providing images in all major formats such as raw, png, jpeg, gif, bmp, wsq
Interoperability with various extractors (client side)
• Any MINEX compatible extractors
Compatibility with various devices
• Support any device in X86 architecture with Windows / Linux OS
Confidential
Support for Iris• Can be supplied through our partner Smartsensor Ltd if required
Sample code for JAVA functions related to Verification, Identification, extractor, Quality test, segmentation. • Currently supports only C++, .NET, but JAVA interface can be supplied if
required
Support for various image formats in the ISO FIR during feature extraction. • Support raw, wsq, png, jpeg, bmp, gif
Confidential
Compliance to various standards. • Minex (ANSI/NIST 378) Compatable, ISO 19794-4
Scaling requirements for 1 Million authentications per hour.
• On an Intel based (>2.6GHz) dual core machine, • assuming template extraction for prob images are performed at client side, • WW matching SDK can handle >32 million authentication (1:1) requests.
Confidential
Summary
• New innovative technical approach – accurate, fast and highly scalable.
• Top tier AFIS accuracy performance -benchmarked by NPL and NIST
• Market leading matching accuracy of low quality fingerprints.
• Technology complementary to other conventional image processing approaches