ADABTS (Automatic Detection of Abnormal Behaviour and Threats in Crowded Spaces)
ADABTS is a four year EU FP7 project addressing the protection of EU citizens, property and infrastructure against threats of terrorism, crime, and riots, by developing techniques for the automatic detection of abnormal threatening human behaviour, while respecting privacy and civil liberties. The automatic systems will support video surveillance operators during their task performance. More specifically, the support system will help the operator to detect abnormal behaviour and threats in crowded spaces, particularly abnormal or undesirable behaviour of individuals and groups of people in crowded spaces, in the context of large scale events, and critical infrastructure.
ADABTS gathers experts in human factors, signal processing, computer vision, and surveillance technology. In the initial stage, focus has been on human factors in order to define and model behaviours. Now, the focus is shifted towards automatic analysis of surveillance data (video and audio). Finally, a demonstration system will be implemented.
ADABTS develop models of behaviours predicting threatening behaviour as it unfolds, thus enabling proactive surveillance. In order to detect behaviour defined by these models, advanced methods for sensor data analysis are needed. These methods extract sensor data features connected to the defined behaviour primitives, and thus detects the presence of potentially threatening behaviour.
ADABTS develops new and adapt existing sensor processing methods and algorithms for detecting and tracking people in complex environments, involving groups of people or crowds. Extracted sensor data features (e.g. tracks, voice pitches, body articulations) need to be related to the behaviour primitives, and moreover they need to be dynamic and adjustable to the context. The algorithms are adapted to run on commercially available low- cost hardware architectures consisting of multi-core CPUs combined with several multi-stream GPUs (Graphical Processing Units). Such hardware is in rapid development by the game industry and represents a huge potential for high performance surveillance systems.
The main impact of the ADABTS project is expected to be on the technological level, with advancements in three directions:
- Understanding of the user needs for automatic detection of abnormal behavior in crowds and new definitions and methods for defining such behavior.
- Methods and algorithms for abnormal behavior detection based on video and acoustic sensors.
- Real time implementation on commercially available low-cost hardware, including an on-line demonstrator at a football stadium.
For more information please contact Asbjørn Berge
.
 |
| Guard is alerted on overview map to threatening human behavior (fighting persons marked in red). |
