Logo multitel

SERKET project – SEcuRisation KEeps Threats

Project presentation: SERKET overview

SERKET is tackling the issue of security for public areas and events by developing an innovative software approach whereby dispersed data coming from discrete devices are automatically correlated and analysed to provide security personnel with the right information at the right time. For this purpose, the project partners aim to design and develop an open platform that will be deployable at low cost.

The project objective was to provide security operators with more knowledge relevant for their missions, to enable them to optimise their decisions. Automatically analysing the information supplied by different types of sensors: videosurveillance cameras, intrusion detectors, access control barriers and microphones.

SERKET detects risk situations; this is the key innovative concept of preventive security: operators can focus on supervision, being alerted by the system as soon as a threat is detected and identified. An operator’s knowledge of the situation is enriched, which is of utmost interest when a situation is particularly sensitive in terms of violence, complex because of the risk of interference with other events and subject to concealed threats. The ideal and inapplicable visual analysis of all images, transferred to the control room when interesting, is then avoided.

In this project, Multitel dealt with the video analysis of crowd scene. The project made use of video and audio data collected during public events, such as the Doudou in Mons.

Doudou in Mons

Doudou in Mons

As illustrated above, an example of Multitel’s contribution is crowd analysis. The objective here is to evaluate crowd density, to detect abnormal movements, object throws, etc. A video analysis algorithm has been developped to keep the position of each pixel in the image, in order to detect the direction and speed of some areas of the image. All trajectories (small lines in the image above) have different colors and lengths. The color is linked to the orientation of the trajectorie and the length to the speed.
Once we have this map of trajectorie, one can easily image to cluster them in order to get the main movements of the crowd, and then to describe the scene.

Related publications

X. Desurmont, R. Sebbe, F. Martin, C. Machy and J.-F. Delaigle. * Performance Evaluation of Frequent Events Detection Systems*. Int. IEEE Workshop on Performance Evaluation of Tracking and Surveillance (PETS). New York, USA. June 2006.

H. Dupuis. * En toute sécurité*. Athena. May 2007.