Intelligent Surveillance
Research Team
Professor Sridha Sridharan (Chief Investigator)
Dr. Clinton Fookes (Chief Investigator)
Professor Massimo Piccardi (Chief Investigator – UTS)
Dr. Simon Denman
Dr. Ruan Lakemond
Jingxin Liu (PhD student)
David Ryan (PhD student)
Ehsan Zare Borzeshi (PhD student – UTS)
Background
As technology continually improves, so too do the applications in which such technology can be utilised. A critical component for managing complex airport systems is the ability to practically assess the state of the system at all times. Intelligent surveillance and automatic information gathering can allow the tracking of persons and objects throughout the terminal, analysis of crowds and queues to identify expected passenger delays, and to also identify anomalous human actions which may indicate threats to the airport and other passengers.
One of the major challenges in these environments remains how to utilise multi-camera networks in order perform the aforementioned tasks. Automatic camera calibration and management techniques are required in order to allow dynamic deployment of cameras across the entire airport infrastructure, and to also make the gathered information accurate and readily available for airport operators.
Figure 1. Example of a virtual gate for counting pedestrian movement through a spatial location.
Research Aims
There are two major aims of the Intelligent Surveillance program:
- To enhance the human component of visual surveillance through the automatic merging of intelligence acquired from large multi-camera networks (which will be visualised within the Airport Information Model); and
- To estimate airport effectiveness and identify important patterns of interest which are directly related to collective enterprise function and emergent properties.
Research Approach
In order to achieve the desired aims, this program will comprise fundamental research into:
(1) Camera calibration and management – automatic techniques to discover camera calibration information and utilise resources within variable and heterogeneous networks;
(2) Person tracking – multi-view tracking research developing hybrid multi-layer motion detectors and scalable condensation filters;
(3) Crowd analysis – research of crowd modelling techniques, business operational measures (such as crowd counting and queue length estimation), and crowd event detection to spawn individual tracks and detect emergent properties in passenger behaviour. This research is particularly useful for Complex Systems understanding and Business Continuity planning;
(4) Human action recognition – development of an ontology to query and flag advanced human actions pertinent to the task of visual surveillance and to promptly detect anomalous behaviours of interest in conjunction with Human Systems research.
Figure 2. Example of automated crowd counting.