1 of 9Image
Research leader Professor Brian Lovell was keen to show off his facial recognition software, currently being trialled at a Brisbane railway station. Lovell has just been recognised (right) by the camera under the screen, which will identify him even if he turns his head up to 60 degrees.
The system that his team is developing focuses on CCTV applications, said Lovell, where challenges typically include using lower-quality images with poor lighting and need completely different algorithms to conventional facial recognition systems, which work on passport quality photos.
"You and I can recognise people from CCTVs but the system can't," he said.
The system his team has developed will identify a person 95 percent of the time in the laboratory.
Faces aren't the only thing NICTA researchers are developing systems to recognise: research engineer Niklas Pettersson's team has created a system that can understand road signs.
Pettersson said the automatic recognition system could be used to update old maps for GPS devices, using small cameras inside garbage trucks or taxis.
The system recognises the signs 90 percent of the time — the same hit rate as human operators.
One challenge has been getting the program to run on a small processor, Pettersson said, so that each camera can send just the coordinates of the sign back to a central system. The information will either be sent via mobile networks or via Wi-Fi hotspots at certain points, such as taxi ranks.
The team is now selecting a camera and processor to build a prototype.
Another traffic project is STAR — Smart Transport and Roads — designed to monitor traffic flow. Instead of laying a wire at traffic lights to track when a car passes an intersection, the project uses video sensors to work out in real time how many cars are standing in a queue.
The real-time information can then be used to allow dynamic traffic control, altering traffic light phases, for example, or permitting more right-hand turns in a lane.
The screen shows a project currently running at a roundabout on the Princes Highway in Albion Park, south of Wollongong. The roundabout has recently had new lights installed, but the lights only turn on when they are needed — when traffic flow becomes heavy and queuing reaches a certain length.
The graphs in the top right-hand corner show how long the queues are in each direction, which is taken into account when determining if lights will go on and for how long.
STAR's efforts received special mention by Broadband Minister Stephen Conroy in his speech at the event, who said it would "lead to more efficient operation of roads", meaning motorists would use less petrol and cut emissions.
Project Brachetto is all about helping teams of teams work together, or "mixed presence collaboration".
This custom-built device is an example of what the team can do: it allows multiple video streams to run on its enormous screen and has a touchscreen on which people can run applications.
"This has every possible thing that you might possibly ever use." Gregor McEwan, research engineer, said.
The team has built an application which can be used as a whiteboard. Different members of a team can draw concepts on a white background by touch. As well as a drawing function, there is also a function for speaking about the diagram. As someone shows connections between objects, a temporary line appears illustrating the body language for those who are looking on from a remote connection. In this image, McEwan is circling an object to emphasise it.
In the SAFE project, alongside facial recognition, the researchers are also working on developing an ad hoc wireless network to operate after disasters such as hurricanes.
This router has a range of around 50 to 100 metres and enables speeds of 20Mbps. It also has multiple antennas transmitting at different frequencies, so several loads can be transmitted at the same time.
The research has also worked on developing a routing protocol that finds a path through the network to minimise interference and achieve the highest capacity possible.
These devices are sensors to be put on farms in areas like the Murray Darling. They sense how much moisture is in the soil, and if irrigation systems need to be turned on and how much water needs to be dispersed into the soil.
The devices, which have an 800-metre range, transmit information to the farmhouse, where information is sent off to a control centre.
In tests that were run over the last two years on four farms in Victoria, the system cut water usage by 20 percent, the team said.