1. Skip to content
  2. Skip to main menu
  3. Skip to more DW sites

Intelligent crowd control

Julian Bohne, HamburgApril 9, 2014

A new "early warning" camera system could anticipate dangerous situations as they arise when large crowds gather. Its developers claim it's anonymous and will protect privacy.

https://p.dw.com/p/1BeDJ
Projekt University of Applied Sciences Hamburg EINSCHRÄNKUNG
Image: Thomas C. Schmidt/SAFEST

Modern society simply wouldn't work without places where large crowds of people gather.

Crowds are a common part of everyday life - at airports and train stations for instance - but also at sporting, music or political events.

But wherever large crowds of people congregate, there's always the potential for disaster. A mass panics can quickly turn a train station, airport and other public space into a death trap.

In 2010, overcrowding led to catastrophe during the Love Parade music festival in the western German city of Duisburg. Twenty-one people died and hundreds more were injured when mass panic broke out.

The event set a precedent for what not to do when organizing large events.

Equally, it was a tragic reminder what can happen when any kind of crowd comes together - whether organized or spontaneous.

A computer counts the number of people within an area based on an infrared image and the body heat signatures of passers-by
Infrared image: a computer counts the number of people within an area based on their body heat signaturesImage: Thomas C. Schmidt/SAFEST

Monitoring crowds

It's inspired a research project at the Hamburg University of Applied Sciences to develop a new kind of crowd monitoring system.

The project is called SAFEST, which stands for Social-Area Framework for Early Security Triggers.

"The central SAFEST scenario is that there's a normal crowd in an area but it gets too large and starts moving irregularly," says Professor Thomas Schmidt, head of the Internet Technologies research group at the university. "SAFEST watches the crowd, counts its movements and people, and gives an alert in the event of any abnormal or dangerous behavior."

Body heat analysis

The system relies on heat-sensitive cameras.

Unlike existing CCTV networks, SAFEST runs autonomously and doesn't require security personnel to constantly monitor the video feeds.

To facilitate this, the researchers have opted for infrared cameras, which detect people from the heat signature their bodies emit.

Professor Thomas Schmidt (left) and PhD student Matthias Wählisch (right) at the Hamburg University of Applied Sciences
Professor Thomas Schmidt (left) and PhD student Matthias Wählisch (right) at Hamburg UniversityImage: Julian Bohne

"The cameras perform a statistical analysis to find out how many people are in a given area, just by measuring the amount of heat," says Schmidt. "So, say there are fifty people with an error of about ten per cent, that's sufficient information to judge whether there's any danger or irregular behavior."

Privacy by design

The researchers say their approach has a number of advantages over existing closed-circuit camera surveillance (CCTV) systems.

Existing CCTV systems generate large amounts of footage, which has to be centrally stored and can be difficult to analyze.

But SAFEST only generates numbers, which considerably reduces the amount of data. It also enables the system to monitor situations in real-time and flag up potential problems.

In addition, the system aims to protect the privacy of those it watches.

Privacy by design: the developers say their system is secure because it only records and analyzes numbers
Privacy by design: the developers say their system is secure because it only records and analyzes numbersImage: Thomas C. Schmidt/SAFEST

"We employ a privacy-by-design approach," explains Matthias Wählisch, a PhD-student working on the project. "We don't record video but infrared images. And the camera only transmits the number of people recorded, not video images."

Marketing advantage

An additional dimension of privacy comes from the fact that facial features are almost impossible to identify from infrared imagery, says Thomas Schmidt.

Schmidt hopes the approach will give his system a marketing edge because it removes the need for legal assessment before it can be installed.

"Deploying privacy-violating systems is complex because you have to involve legal people and all these kinds of things," says Schmidt. "Being privacy-preserving by design allows you to deploy the system at low complexity and rapidly, without any involvement of legal departments or other people."

The Internet of Things

SAFEST uses the concept of the Internet of Things. This means that each individual component of a network uses Internet protocol and is able to transmit data in a standardized way.

In the case of SAFEST, this lets the cameras communicate with each other, as well as with the system and with other sensors, such as smoke detectors.

It also removes the need for human interaction with the system, so it can run autonomously.

Projekt University of Applied Sciences Hamburg EINSCHRÄNKUNG
The basic infrared image reveals few distinguishing feature of the peopleImage: Thomas C. Schmidt/SAFEST

But the researchers want to simplify communication in the system further still.

They are developing an operating system that is specifically geared towards the Internet of Things.

It's called RIOT and the researchers hope it will make integrating new devices into existing networks seamless.

"You can think of applications running on this operating system, in conjunction with the Internet of Things, as apps on an iPhone or applications on a PC," says Schmidt. "They are small programs that perform certain tasks, and they run and interact only with our operating system. So if you move the application to a different device with the same operating system, it will simply run."

The basic functionality of SAFEST is already there. But it will be a while yet before the system is market ready.

At the moment the researchers are trying to make the components smaller and more robust. The aim is to reduce energy costs while at the same time increasing reliability.