HomeNational Research MapsLogosContactAdmin Login

Login

Insert
Update

Search
Show all
Show All

Project Idea

OrganisationArtificial Vision and Real Time Systems Laboratory
Inserted:2016-03-30
Project TitlePrevention and detection of security threats in public environments
H2020 Topic ListSEC-12-FCT-2016-2017 Technologies for prevention, investigation, and mitigation in the context of fight against crime and terrorism
Role within the Consortium* Project Partner
Type of activity* Research
Project DescriptionThe term Homeland Security expresses a broad concept with a precise objective: the protection / safety of persons and goods (infrastructure, key resources, etc..) in a given territorial area. This is achieved through the protection of the territory from threats, terrorist attacks, organized crime and urban, adverse natural events, malfunctions or errors related to the functioning of infrastructures having large importance for security.
In modern urban areas there has recently been a continuous installation of sensors for the control of the territory, particularly visible and thermal cameras. This requirement is motivated by the need to increase the safety and life quality of the citizen. The current situation is characterized by installations of sensor networks poorly cooperative whose functioning is limited almost exclusively to the human interpretation of monitored scenes in the total absence of feedback and control / optimization of the data acquisition process. The backwardness of the technology used today and against the large amount of data / information that sensor networks generate do not allow the current security systems to meet the demands of homeland security.
The proposed project thus aims to improve the research and development of novel algorithms and technologies to optimize the process of automatic video data acquisition, integration and interpretation using the sensor networks available in public environments. The final goal of the project is to develop a system for the management of large sensor networks, whose manual monitoring by human operators is typically sub-optimal. More specifically, we will focus on possibly crowded public areas with well-defined entrance points (gates), such as train stations, airports etc. This scenario allows to deal both with security-oriented prevention and detection tasks. Prevention is done at the gates, by acquiring biometric tags to notify the presence of known, potentially dangerous individuals that can be later tracked across the environment using sensor reconfiguration and target re-identification techniques. Detection of anomalous or dangerous events instead is performed in the entire environment: pre-processing modules help the system to focus its attention on the most salient areas (given a proper definition of saliency) where explicit event detection algorithms are applied. This analysis step is explicitly studied to work even in the case of crowded environments, where the detection of specific individuals is impossible.