Advances in Iris Recognition Interoperable Iris Recognition systems Date 5/5/09
Agenda How best to meet operational requirements Historical Overview of iris technology The current standard Market and Technological drivers Operational programs using the ISO standard Next steps in interoperability Research IREX 2008 Quality assessment as an independent function DeFacto Standard Templates from 2Pi Algorithim * IOM is a register Trademark of Sarnoff Corporation 2
Iris Recognition The Biology and The Technology 1. Iris is NOT the Retina 3. 512 Byte Digital Template 2. Digital Image/ Subtle Light 4. Authentication Robustness
Salient Points Iris Recognition Then The first commercial iris products appeared in late 1997. Limited set of formidable players for the first 10 years. Single algorithm vs. multiple algorithms in the market. Restricted availability of core technology camera know how. More spent on litigation than collected revenue. Now Many new entrants in the past 24 months for cameras and matching algorithms. Significant R&D in government, academia and commercially. 4
Market Drivers General Need to process larger populations at a higher speed Minimize the probable outlier population using multiple biometrics The relative success of iris in theater Adoption of iris into the national criminal database at FBI Technological Iris at a distance Iris in motion Portable iris Less than pristine iris image recognition Outdoor use 5
Iris Algorithms Historically the segmentation and template generation process has been a singular function. Image Quality assessment routines have been an integrated part of the template generation process Imposter distribution curves have not traditionally been made available to the research an integrator community There have been a limited number of commercial offerings Segmented Image 6
Interoperable Iris Recognition Definition ISO Iris Standard 19794 6 Rectilinear and Polar (segmented & un segmented) Image definitions Basic quality metrics for interoperable systems are defined A basic challenge is the size of the data as defined in the specification 300 KB 16 KB
Image size reduction Advances in the segmentation process may allow smaller images Polar image 8
Standards Activity surrounding Iris IREX 2008 The IREX 2008 test is designed to measure the accuracy, interoperability of rectilinear and polar iris images formatted according to the ISO/IEC 19794 6 standard, under JPEG and JPEG 2000 compression. The test itself will proceed in summer 2008, with a view toward making results available ahead of the SC 37 meeting in January 2009. Commercial Imager Evaluations (NIST) Sponsor FBI Multiple Biometric Grand Challenge (NIST) Low quality still images High and low quality video imagery Face and iris images taken under varying illumination conditions Off angle or occluded images FBI NGI 9
Programs Using the Iris Standard DOD Field operations with portable iris capture and recognition device Images/Templates are exported to the BAT 4.0 field repository Matching using a proprietary algorithm US Registered Traveler CIMS uses rectilinear image for de duplication Service providers use compressed polar image on a smart card for verification at the airport kiosk. Some international RT programs (Privium & Saphire ) use templates Border Crossings Jordan, Oman,Qtar, Saudi Arabia, UAE Privium & Saphire ( use proprietary template on the smart card) CAN PASS (AKA Nexis Air) 10
NIST IREX 08 IREX 08 is going to test three different compact formats of iris images: cropped images, Un segmented polar images, ROI (region of interest) images. NIST asked vendors to submit software that makes the compact formats and matches templates. But we were not required to support all three image formats. We can either support NIST will also test the boundary of various algorithm's performance characteristics with varied degrees of compression on iris imagesimages 11
Images from a distance Image matching probe vs. gallery The image at the right successfully matches against itself and the pristine quality enrollment image. Different illumination A successful recognition image captured at 2 meters The image at the right matches against itself, but not to an image with standard illumination. 12
icap Iris Capture & Analysis Platform icap Flex 2 idata ISO SDK Image Processing Library Windows PC Iris images SQL Database June 2008 LG Electronics USA Iris Technology Division
Why icap? Some Stage One Objectives Provide a easy to use tool to enable the study of iris images from various commercial and prototype imagers. Process images of lesser than perfect quality as compared those used in commercial deployments of iris recognition technology which have traditionally expected pristine quality images. Initial GUI targeted at the mid level analyst which may have limited experience with iris recognition technology. Prepare a framework for the addition of other available iris matching algorithms. Work towards the separation of image quality assessment, segmentation, and template generation as discrete operation steps. 14
Different Illumination Commercial iris imagers use illumination in the IR bad from about 700 nm to 850 NM. IARPA Research Program Use other wavelengths Diffused 1550 NM Iris illuminated with 1550 NM Iris illuminated with 950 NM 15
Thank you! Tim Meyerhoff LG Electronics U.S.A. Inc. Meyerhoff@lgiris.com 908-803-4596 16