Motivated by security concerns, an emerging class of digital video cameras provides unprecedented ability to zoom in and capture high-resolution video images within a wide field of view. This capability is desirable for practical applications, such as security, public viewing of progress at a construction site and in research applications such as monitoring behavior in public sites.
However, such cameras raise significant privacy concerns. We propose to investigate a new approach to providing some measure of visual privacy by hiding an individual's identity while allowing observation of their physical actions and other live images of the scene. Our approach is to utilize wearable "markers" that can be detected by image processing software in real time. For example, we could provide inexpensive hats of a particular color or pattern at the border of the space where the camera is present (similar to the respectful hats or leg-coverings that are made available at the entrance to churches or synagogues). Our objective is to develop "respectful cameras."
1. In real time (30 frames per second) it is possible to reliably
identify moving markers worn by individuals in a video stream and to overlay obstructions over faces/bodies to hide the identity of individuals.
2. The majority of people who desire visual privacy will find this
method provides a reasonable means and sufficient level of visual privacy.
Respectful Cameras: Detecting Visual Markers in Real-Time to Address Privacy Concerns. Jeremy Schiff, Marci Meingast, Deirdre K. Mulligan, Shankar Sastry, and Ken Goldberg. International Conference on Intelligent Robots and Systems (IROS) . [Submitted] San Diego, California. October 2007 [1.2MB .pdf].
Respectful Cameras. Jeremy Schiff, Ken Goldberg, Marci Meingast, Deirdre Mulliganm Pam Samuelson. Exploring the Privacy Implications of Trustworthy Information Systems Workshop for TRUST researchers. University of California, Berkeley. October 5-6. [4.4MB .ppt]
Computer Vision Class Paper [3.2MB .pdf]
An overview of our initial solution [66KB .pdf]
[3/28/07] Test evaluating our Dynamic Marker Tracker (with particle filter) at the CITRIS construction site [4.0MB .wmv]
[3/28/07] Test evaluating our Static Marker Detector (no particle filter) at the CITRIS construction site [3.6MB .wmv]
[3/28/07] Test evaluating our Dynamic Marker Tracker at CITRIS with only RGB colorspace [4.5MB .wmv]
[3/28/07] Test evaluating our Dynamic Marker Tracker at CITRIS with only HSV colorspace [4.0MB .wmv]
[3/28/07] Test evaluating our Dynamic Marker Tracker at CITRIS with only LAB colorspace [4.1MB .wmv]
[3/28/07] Test evaluating our Dynamic Marker Tracker at when two people cross in our Lab [4.2MB .wmv]
[3/28/07] Test evaluating our Dynamic Marker Tracker under changing lighting conditions, in particular a bright light being directed at the visual marker causing an extreme specularity [1.9MB .wmv]
[7/27/06] Outdoor test moving between light and shade, illustrating just hat detection on each frame independently [5.6MB .avi].
[7/27/06] Outdoor test moving between light and shade, with Baldassari privacy dot where each frame is processed independently [8.8MB .avi].
NSF Science and Technology Center, Team for Research in Ubiquitous Secure Technologies, NSF CCF-0424422, with additional support from Cisco, HP, IBM, Intel, Microsoft, Symmantec, Telecom Italia and United Technologies.
Respectful Cameras and John Baldessari (video), in Current TV by Tom Green (Nov. 2008)
Surveillance Cameras that Obscure Faces, by Bruce Schneier, (June 2007)
Respectful Cameras, in MIT Technology Review by Brendan Borrell (May 2007)
Spy camera hides friendly faces, in Personal Computer World by Martin Lynch (May 2007)
Cámaras que resguardan la privacidad
Private Eyes, in California Magazine by Amy Goldwitz (Oct. 2006)
University of Kentucky's Multimedia Information Analysis Lab
The How to Disappear kit is a counter-reaction to the ever increasing surveillance of citizens.
Related Visual Privacy Work at the University of Kentucky
Proprietary Product for Visual Privacy Enforcement
A security camera company: Pixim
News article about automated blinding of CCTV cameras with Infrared LEDs.
Article summarizing some recent approaches to providing security with privacy, including Google's new addition of face detection and blurring in Street View for the area of Manhattan.