Usability Considerations for Face Image Capture at U.S. Ports of Entry NIST International Workshop o

1
Usability Considerations for Face Image Capture
at U.S. Ports of EntryNIST International
Workshop on Usability and Biometrics June 23-24,
2008 Lawrence D. Nadel, Ph.DNoblis
2
Agenda

• US-VISIT background
• Potential face recognition applications in
  US-VISIT
• Port of entry (POE) operational environments
• Usability considerations

3
US-VISIT Background

• US-VISIT provides biometric identification and
  analysis services to agencies throughout the
  immigration and border management, law
  enforcement and intelligence communities
• US-VISITs services help decision makers
  accurately identify people and assess whether
  they pose a risk to the United States
• Biometrics captured from non-US citizens, ages
  14-79
• Biometrics captured at Entry enhances security
  and facilitates legitimate travel
• Fingerprints
• 10P enrollment/background check, four-finger
  verification
• Exit verification (under development)
• Facial image
• Human verifiable traveler history
• Currently no automated face recognition
• Biometrics captured at Exit (to be determined)
  identify overstays, crosscheck watch lists

4
Applications for Face Imaging/Recognition

• Human comparison
• Compare live photo with visa and/or past photos
  for visitors who have not been fingerprinted
• Compare live photo with e-Passport photo for
  first time Visa Waiver Program visitors
• Increase verification confidence through decision
  level or score level fusion
• Supplement fingerprint check for detecting
  aliased (duplicate) records or fraud
• Search face-only watch lists

5
Diversity of Operational Environments,Use Cases,
and Travelers

• POE types
• Land pedestrian, car, truck, bus
• Air small plane, jumbo jet
• Sea small boat, cruise ship
• Ambient environment - indoor/outdoor
• Variable illumination - day/night, directional,
  multi-spectral
• Entry formal inspection stations
• Exit little or no inspection infrastructure
• Travelers
• Cooperative, non-cooperative, uncooperative
• Multiple languages, cultures, appearances/clothing
  
• Hands full - luggage, packages, small children

6
Air POE Environment

• Key factors for face recognition
• Pose angle
• Interocular pixel resolution
• Illumination
• Subject distance from camera
• (head size, distortion)
• Background

7
Land POE Examples
Pedestrian Exit
Entry
8
Usability-Related Interactions
Workstation/ System
Traveler
CBP Officer
9
Cooperative Traveler

• Indicate that a picture is being takenwhere and
  when
• Image capture sensor should look and sound like a
  camera
• Provide simple and clear guidance (oral/written,
  foreign language, still images, video)
• Limit physical degrees of freedom, e.g., indicate
  where feet should be placed on floor to control
  distance to camera
• Accommodate traveler whose hands may be occupied,
  e.g., baggage, small child
• Align camera with users faceaccommodate
  variable height (short/tall, standing,
  wheelchair) multiple cameras, portrait mode,
  wide field of view digital camera

10
Non-cooperative Traveler

• Human factors engineering to direct travelers
  gaze at camera and have traveler pause for photo
  no conscious effort on part of traveler
  required
• Printed signage
• Video display (static or variable)
• Strategic chokepoint
• Top of escalator
• Turnstile
• Portal

For US-VISIT Exit
11
CBP Officer

• Officer needs to review documents, operate
  workstation, and interview and observe traveler.
  Position equipment to minimize officer movement
  and minimize use of peripheral vision.
• Simple and logical workstation GUI.
• For officer positioning of camera, show geometric
  overlay on video screen to indicate proper
  placement and size of image to be captured.
  Provide easy to use automated or manual camera
  control.

12
Workstation/System

• Audio-visual feedback to both traveler and
  officer
• Automated image capture (quality in the loop)
• Automated control of camera focus, exposure
• Electronic control of pan-tilt-zoom (PTZ) camera
• Automated face finding and quality assessment
  algorithms
• Select image
• Crop

13
Quality-in-the-Loop Face Image Quality
Improvement and Face Recognition Study

• Select and assess representative cameras
• Webcams
• Video with Pan-Tilt-Zoom
• Digital Still
• Wide Dynamic Range

• Select and assess several face quality metric
  software tools
• Inter-eye distance, head position, face contrast,
  lighting uniformity,

• Integrate selected cameras capture photos and
  video streams run image quality software
  post-capture to assess impact

• Determine best hardware/software combination
  integrate to run real time assess potential
  impact on image selection

14
Simulated Demonstration of Quality in the Loop
for Image Selection (Webcam)
(0-10)
(sec.)
15
Discussion

• Lawrence D. Nadel, Ph.D.
• Phone [phone]
• Email nadel_at_noblis.org