blog

Here in London, we have seen our fair share of hooligans (sorry – ‘rioters’) in recent times. The London riots seemingly started with a gathering of people in peaceful protest which, at some point, escalated into rioting. Such instances occur in areas where there are a large number of people gathered together, such as in sporting events and parades and the key to preventing these is early detection and dealing with hostile or aggressive behaviour before the situation escalates.

(from https://www.tunisia-live.net/2011/08/07/after-riots-and-peaceful-protests-relative-calm-returns-to-jebiniana/)

Dr Dominik Endres from the Hertie Institute for Clinical Brain Research and Center for Integrative Neuroscience in Tübingen, Germany, investigated the search strategies of those whose job it is to survey large crowds, as part of their research into the development of a computational surveillance system that can mimic visual processing in the brain. “To that end, we investigated how human experts search for security-relevant events, such as mass panic, fights etc.”, says Endres.

In work presented at the European Conference on Visual Perception 2011, Enders and his colleagues investigated the effects of spatial and temporal expert knowledge on detection of salient events at a soccer stadium. Results suggested that experts use scene-specific priors that depend on the spatial layout of the arena and different time phases of the game. Surprisingly they also found that, when spatial and temporal priors are removed, the gaze patterns of naïve and expert observers are quite similar in the first few seconds; however, expert observers were much better than naïve ones at both event detection and classification. “In other words, experts and naïve observers look in a similar fashion, but experts see more”, Enders explains.

The insights gained from studying the gaze patterns of those with many years of experience in looking for possible security-related events is currently being implemented in an automated crowd-monitoring system, as a computational cognitive model of visual processing, which should help ensure the safety of visitors at mass events.

If it makes for more peaceful protest, then I’m all for it!