Project Description

Project Aims

SUBITO will develop a multi-layered processing system capable of utilising a range of sensors to detect abandoned objects, assess the threat and provide preand post event tracking of their owners.

Working closely with the end users, the team will design a system that is capable of distinguishing between genuine threats and false alarms in order to direct the focus of attention of the user to situations of the highest priority.

The SUBITO programme will deliver a demonstration of semi-automated data processing built upon existing surveillance technology that will provide automated realtime detection of goods that have been abandoned, the fast identification of the individual who left the goods and the fast determination of the current location of that individual or his/her followed path.

This demonstration will be achieved using existing infrastructure and security technologies from real locations under standard operational conditions, as well as using staged scenarios.

It will also show that this user-focussed result is achieved through a generic approach that can be applied to solve analogous problems in diverse applications.

The aims of the program will be achieved by building upon the research carried out to date in this field, in particular the EU Framework 6 project ISCAPS (SEC4-PR-013800).

Scientific Objectives

The following scientific objectives will be pursued by the project:

Robust Detection - SUBITO will review and develop novel extensions to this work to develop a stereo based solution to lighting invariant background subtraction for the task of robust abandoned object detection.

Robust and Long Term Tracking - the project will investigate how these problems can be overcome through novel fusion of the outputs of independently run tracking algorithms.

Robust Identification - In SUBITO, robust categorisation and identification of objects will be achieved through fusion of a number of visual cues including size, gait, shape and appearance.

Behavioural Analysis - The project will test the probabilistic models and logic to increase the discriminating power of present day behavioural analysis systems.

Facial Recognition for multi camera surveillance application - The application and advancement of face recognition technology in this field as well as the integration with further complex software components (detection of unattended goods, identification of the owner) are important steps beyond the state of the art.

Sensor Fusion - State of the art probabilistic methods will be explored as part of SUBITO as well as fusing pattern classifiers.

PTZ Cameras -Automatic and robust detection of abandoned goods using a camera network is a complex task as the system is observing a wide 3D scene including occlusions and changes of appearance.

Performance Evaluation - In SUBITO algorithmic robustness will address the development of a novel evaluation methodology (based on appropriately chosen benchmarking criteria) to demonstrate improved object detection, tracking, classification and action recognition capability