To set the scene for the making session of the workshop we gave two short presentations to participants. The first was on existing strategies of urban camouflage. This was followed by an introduction to computer tracking systems.
Introduction to computer tracking systems: The challenges of computer vision pedestrian tracking is an active field of research with several decades of progress. The seamingley easy task for humans of identifying and tracking a moving object using vision is complicated for computers in real-world environments by changing lighting, object shape, texture, direction, line of sight, etc.
Early successful surveillance programs relied on calculations of the changes in pixel colour from one video frame to the next to identify 'blobs' of movement that could be a pedestrian. Adapting to changing environmental conditions, keeping track of individuals over time and improving blob selection all helped to make such tracking suitable for outdoor use. However, robustness of tracking and the related challenges of identification of a particular person through unique 'biometric' characteristics, provide ongoing challenges for the designers of these systems.
An understanding of the improvements that have been made to pedestrian tracking provides potential actions that can be taken to control personal visibility under the gaze of modern computer vision. We looked at a range of methods implemented to improve robustness.
Pedestrian movement can be modelled, so that the shapes can be classified.
Textures of clothing and patterns of the human body can be searched out and learnt through computer vision training.
Key features found in clothing, face and appearance can be approximated as trackable patterns.
Changes in weather and lighting, and occlusion from foreground patterns can be adapted out to remove distraction.
Crossing paths and groups of people together can be tracked through flocking agents, or simulated with simple physics of momentum.
Walking rhythms can be identified through gait analysis.
Models of 'normality' for people in a particular space can be built up over time, triggering digital trip-wires when a movement or path is awry.
Multiple view points can be used to build a three dimensional map of a space.
If it is possible to control an environment more, a space can be actively changed to aid tracking, through both visible and infra-red lighting.
Download the slides from this presentation here (PDF)
Introduction to urban camouflage: This presentation illustrated how people and objects can be disguised and camouflaged so as to be visually deceptive. The images from the presentation are below.
Blur building by Diller & Scofidio
Zebra by Desiree Palmen
Urban Camouflage by Aya Tsukioka
Madrid Street Gallery
Hair clip by Humans-Since-1982
Crocheted facial features by Howie Woo
Too Beautiful to be True by Meike Harde
Britannica Colour Chart for dazzle camouflage
HMT Olympic painted in dazzle camouflage
Electro Draught Excluder by Dunne & Raby
Growth of Surveillance in Public Space by Applied Autonomy
CV Dazzle by Adam Harvey
WOW Laguna workshop by Aram Bartholl