visual monitoring of people in private spaces
TRANSCRIPT
Francisco Flórez-RevueltaInterdisciplinary Hub for the Study of Health and Age-related conditions (IhSHA)
Visual monitoring of people in private spaces
Visual monitoring in publicspaces
Monitoring in private spaces
Cameras in private spaces
Fabien, C., Deepayan, B., Charith, A., & Mark, S. (2011). Video based technology forambient assisted living: A review of the literature. Journal of Ambient Intelligence andSmart Environments (JAISE), 3(3), 253-269
Chaaraoui, A. A., Climent-Pérez, P., & Flórez-Revuelta, F. (2012). A review on visiontechniques applied to human behaviour analysis for ambient-assisted living. ExpertSystems with Applications, 39(12), 10873-10888
Computer vision in AAL
Fall detection and prevention
Identification of Activities of DailyLiving
Behaviour monitoring
Physiological monitoring
Person-environment interaction
Rehabilitation
Support to people with cognitiveimpairment
Cons:Cluttered environmentsOcclusionsPrivacy preservationLimited field of view
Pros:Richer informationTracking of coarse and fine movements/actionsSynergies with other servicesEase to interpret
Pros and cons
Idea behind
Idea behind
Appropriate measures need to be considered to maintain privacy and increaseacceptance
The notion of privacy is highly subjective. It depends on theindividual.
Several factors are involved:The private “thing”The observer
An image conveys the following information about an individual:Identity (Who?)Appearance (How?)Activity / Behaviour / Event (What?)Time (When?)Location (Where?)
Privacy preservation
Image Redaction: Modify an image or a sequence of images so as to protectobjects (visual clues) appearing on them
But…The image must retain its utility
A trade-off between privacy protection and image utility is needed
Privacy must be adaptable to the individual
Ensuring visual privacy
We propose a privacy protection scheme that is aware of the context
A set of redaction methods is used
A context describes “any” situation:Identity of the observerIdentity of the observed person (to retrieve the privacy profile)Closeness between person and observer, e.g. relative, doctor, neighbourAppearance (dressed?)Location, e.g. kitchen, living roomEvent, e.g. cooking, watching TV, fall
Users provide their privacy preferences by linking instances of the context withprotection/visualisation methods
Privacy by context
Privacy by context
Visualisation levels
Original Pixelate Blur Emboss Silhouette
Skeleton Avatar Invisibility
Visualisation levels
Tests with RGB-D cameras
Tests with RGB cameras
Improvements in the calculation of the contextIdentity (Who?)Appearance (How?)Activity / Behaviour / Event (What?)Time (When?)Location (Where?)
Improvements in foreground/person detection
Most of activity/behaviour identification systems are validated in labs, not in realenvironments
Privacy of the environment
Main issues: current and futurework
Padilla-López, J.R.; Chaaraoui, A.A.; Gu, F.; Flórez-Revuelta, F. (2015). Visualprivacy by context: proposal and evaluation of a level-based visualisationscheme. Sensors, 15(6):12959-12982.
Padilla-López, J.R.; Chaaraoui, A.A.; Flórez-Revuelta, F. (2015). Visual privacyprotection methods: A survey. Expert Systems With Applications, 42(9): 4177-4195.
Flórez-Revuelta, F.; Gu. F.; Pierscionek, B.; Remagnino, P. (2015). White paperon AAL systems and associated privacy issues. Public report, BREATHEConsortium.
Padilla-López, J.R.; Flórez-Revuelta, F.; Monekosso, D.N.; Remagnino, P.(2012). The “Good” Brother: Monitoring People Activity in Private Spaces. InDistributed Computing and Artificial Intelligence, pp. 49-56, Springer.
http://www.breathe-project.eu/
More information
Francisco (Paco) Fló[email protected] @fflorezrevueltastaffnet.kingston.ac.uk/~ku48824 franciscoflorezrevuelta