Dynamic Crowd Measurement - CNN

Three benefits of using AI for crowd analysis

Thursday 02 Aug 2018

The DCM team recently completed a round of testing & validation of our Artificial Intelligence (AI) based crowd decision solution at Australia’s largest outdoor light show event. In our journey towards market release, we have conducted many deployments of the solution at large events to refine both AI elements and our new blueprint for crowd decision analysis.

 

Dynamic Crowd Measurement - Setup
Here’s our Retrospective:

1. The confidence of improved decision making:

Police reported feeling more confident making decisions when backed by the crowd data because the accuracy of the software can be proven, the dashboard shows the crowd image as well as numbers which are both retained as evidence of action. Previously human subjective information was varied and unsubstantiated. We are not saying intuition is not relevant, however it is improved with real-time data and analytics. The cameras are scanning hundreds of thousands of data points per minute.

 2. Perception is not reality:

DCM can run through most existing cameras, ideally mounted above the crowds in vantage points, and as a result this is significantly better than trying to estimate crowd numbers from a ground perspective.

We compared event personnel estimates of crowd congestions and therefore decisions on when to make changes with the real-time data analytics across the event footprint. This comparison highlights variances; in some cases even experienced personnel judgements on crowd density were at 80 % variances.

 3. Coupling AI and alerts bring improved decisions:

The DCM dashboard of alerts assisted event personnel to discuss scenarios and options for crowd flow decisions and strategies. Our crowd flippers (crowd flow pressure diversions to provide relief) based on real-time data meant that decision paths could be implemented almost immediately.

There is much to be gained not only from the public safety of crowds at the event but also benefits for front-line event personnel who need to make real-time decisions. Here are the areas we have highlighted to improve as we iterate our solution framing:

  1. Importance of training of event staff with a new way of managing and making decisions. Especially in cases where crowd flows contain crowd pulses from transport hubs disembarking and natural concertina effects, early detection of these signals can smooth mini-jams.
  2. Imagery can be recorded for training use and crowd flow design purposes and future modelling
  3. Detection allows new techniques for crowd diversion and crowd calming, for example ‘flippers’ can be set up to relieve pressure from the main crowd flows when the pressure becomes too high. This is not effective without pre-warning by detection of early signals, previously we waited until a crush occurrence before we reacted because we couldn’t sense increasing pressure quantitatively.

If you would like to see DCM and how our new framework can be used to keep on top of your crowd analysis and planning, please request a demo, and we will be happy to assist.

 

Dynamic Crowd Measurement - Crowd Density