Main Objectives

The project's main goals are to help preventing, to detect and to support the investigation of crimes committed in bank foyers, where the special focus of tripleB-ID lies on the prevention and on the support of investigation.

Behaviours of interest

Together with the project partners, several scenarios were defined as suspicious. These scenarios include amongst others:

• wandering around in the foyer without using a machine (e.g. the ATM) or contacting a member of staff over an extended period of time.
• operating a machine over an unusual long period of time.
• more than one person operating a machine at the same time.

In addition to these suspicious activities, we are interested in detecting aggressive behaviour. Beside these well defined scenarios, we also intend to detect unusual behaviour, i.e. behaviour that cannot be described by the data previously labelled as normal.
The detection of suspicious behaviour may have several implications, such as informing a member of staff to judge the situation and take appropriate action or labelling the video data as suspicious. Video sequences labelled as suspicious can subsequently be stored for a longer period of time.

Forensics

In case a crime cannot be prevented, the system should be able to provide information usable by the police, such as a detailed high-quality view of the face and an accurate estimation of the size of a suspect.
In order to collect higher quality facial images, standard face detection is performed on the static camera. The dynamic camera is subsequently zoomed to the detected area and collects facial images. The face detection is coupled to a pedestrian detector to reduce the number of false positives. The higher resolution facial images are then connected to the trajectory of the person.
The height of a person is estimated by applying visual metrology. The height measurement of a person is updated and consolidated during the tracking. Furthermore, an interface for performing manual measurements is provided as well.