In 1999 Sentient, a Melbourne based company, established the goal of developing a spatially-aware vision system suitable for autonomous robots.

The company named the technology the Sentient Vision System (SVS), and by early 2004 Defence recognised the potential of the product.

One of Sentient's founders, Dr Paul Boxer, was reading a Defence Science and Technology Organisation (DSTO) produced magazine that highlighted a similar technology DSTO was developing.

Dr Boxer noted that the technology aimed to perform similar functions to SVS but performed the tasks in much slower time.

‘We had created the SVS technology for a different purpose but it was capable of performing the Video Motion Target Identification (VMTI) task that Defence was seeking technology for, in real time,’ Dr Boxer said.

At about the same time DSTO contacted Sentient to purchase a commercial surveillance product and learned of the innovative SVS VMTI capable technology.

The result is a current Capability and Technology Demonstrator (CTD) project, which began in October 2006, that aims to demonstrate a high definition, real-time VMTI system running on a personal computer that will locate and track small moving ground targets from the air.

No other system is presently available that can do this effectively and as a result, the CTD program has generated interest in the Sentient technology from several international prime defence companies.

Under the CTD Sentient has developed its original SVS technology into a new product, named ‘Kestrel’.

Dr Boxer said the original SVS technology had a few limitations that Sentient overcame through the CTD.

‘The older technology assumed the surface that was viewed was flat, without major obstacles such as trees and buildings. And, you could not process large images in real time,’ he said.

‘Through the CTD we have developed Kestrel, a fast and accurate technology that can process up to 6 mega pixel images in real time and find targets that even humans couldn’t see. And, it is possibly the first computer vision system capable of VMTI in real-time on a low-specification Pentium 4 personal computer,’ Dr Boxer said.

Its primary task is land area surveillance for situational awareness and it is able to detect car-sized moving objects from one second of video imagery, covering 16 square kilometres.

Areas where Kestrel may prove to have a capability include:

• passive ‘look down’ detection of aircraft and land vehicles from high altitude UAVs;
• search and rescue operations in environments ranging from heavily wooded, mountainous terrain to the maritime environment;
• wide area surveillance to detect activity, evidence of recent activity and landscape change, such as trespassing and remote covert operations;
• border patrol;
• passive guidance for the terminal phase of precision guided munitions onto moving targets;
• high speed generation of large mosaics of landscape imagery; and
• autonomous, accurate cross-cueing of on-board sensors.

Dr Boxer said the technology is differentiated from other available systems because it discards traditional computer vision methods.

‘We have developed new vision methods based on biological systems. The technology is designed to emulate the way in which an animal would see,’ Dr Boxer said.

‘Where older systems tried to work out what things were by their shape our system looks at motion, the way things move,’ he said.

A suite of Kestrel products is likely to be on the market when the CTD concludes in mid 2008.

Dr Boxer commended the CTD program. ‘The entire process runs really smoothly and is a benchmark by which other research should be funded. Without acceptance into the CTD program this technology would not have been developed to the level that it has been,’ he said.

Using the Kestral system, a unmanned aerial vehicle, such as this Aerosonde, is capable of detecting and tracking both vehicles on the surface and low flying aircraft, even though they only appear as tiny points on the landscape.