The research is conducted by an international consortium consisting of FOI (SE), BAE Systems (UK), Detec (NO), Home Office Scientific Development Branch (UK), Institute of Psychology – Ministry of the Interior (BG), SINTEF (NO), University of Amsterdam (NL) and TNO (NL).

For this part of the study the consortium spoke with potential users of the envisioned system and inventoried their requirements and constraints. Using questionnaires and interviews and a literature survey, ADABTS determined some of the most relevant domains and promising scenarios within these domains. These scenarios will be used to generate lists of detectable elemental behaviours that together, and in context, may indicate a potential threat.

One of the findings of the survey is that in many cases operators can only concentrate for up to 40 minutes, after which “video blindness” sets in and detection rates drop severely. In contrast, it is quite common that operators in control rooms are required to observe the situation for 6 to 12 hours. The survey also shows that the number of security cameras that has to be overviewed can be extremely large, whereas the number of monitors that can feasibly be handled by a single person ranges between 4 and 16. This leads to a rather ineffective use of camera surveillance equipment (a study showed that only a fraction of the incidents are detected in real-time (pro-actively) in shopping mall/city centre areas). Such findings support the idea that a support system that filters out the relevant information (such as developed within ADABTS) can be extremely useful.

Sound is still largely unused in current camera surveillance systems. However, the survey indicates that sound (which will be exploited by the ADABTS system) is considered to be a valuable contribution to camera surveillance, although the implementation may be less straightforward than video, e.g. due to privacy issues. It opens up various interesting possibilities such as event detection from sound (plus combination with video), sound localization (which can be used for steering the cameras), and directional sound enhancement. All possibilities are being investigated within ADABTS.

It is recognised that the detection algorithms are and will be (for a long time) far from perfect. This means that they will generate a large number of false alarms. This is enhanced by the effect that incidents are typically rare, so even high performing detection algorithms will generate a relatively large number of false alarms. This can make the algorithms literally useless: in cases in which the number of false alarms was too large people were found to switch off the support system, an effect known as the “cry-wolf effect”. This highlights the need for a Human-Machine Interface (HMI) that makes effective use of the system and operator capabilities. The ADABTS report contains some ideas on how this issue could be dealt with.