University of Wollongong to receive grant to develop AI for Geospatial Intelligence

As part of the $1.2 Million Critical Intelligence Program, the University of Wollongong has been successful in its application for second-round funding to develop a machine learning (ML) tool for detecting and identifying maritime vessels using satellite imagery. The funding is part of the Australian Geospatial-Intelligence Organisations (AGO) Analytics Labs Program.

Associate Professor Son Lam Phung, Senior Professor Salim Bouzerdoum, and Dr Fok Hing Chi Tivive, from UOW’s School of Electrical, Computer and Telecommunications Engineering, will lead the project – What Vessel Is That? – to develop a machine learning tool that can detect and classify ocean vessels from satellite synthetic-aperture radar (SAR) imagery.

The idea of this project is to develop a tool that can detect and quickly differentiate a fishing boat from a patrol boat, for example, or a military ship from a cargo vessel. It is hoped that the system will be able to use commercial SAR systems to undertake automated identification of vessel categories.

“There has been an increasing interest in maritime surveillance to counter illegal fishing and maintain maritime rights and interests,” Professor Phung said.

The project will also involve postdoctoral researcher Thanh Le and several PhD students (including Hieu Phan and Ly Bui). Hence, it is also helping to train the next generation of researchers with skills in Artificial Intelligence (AI) and ML to address defence and security needs.

The AGO’s Analytics Lab Program (AGO Labs - which is being coordinated through FrontierSI, a not-for-profit organisation specialising in expanding the use of spatial mapping, infrastructures, positioning, geodesy, analytics and standards) has just completed its second round of funding. The program is designed to engage industry better and address the capability challenges that the AGO face through industry projects.  

UOW were one of three organisations to receive part of the funding, with the Minister for Defence and Industry, Melissa Price congratulating the successful applicants, saying their proposals demonstrated the immense value that could be realised when Defence and Australia’s defence industry team up to overcome key challenges. 

The successful organisations have each been awarded six-month contracts to deliver rapid insights to AGO. The participants will work closely with AGO subject matter experts to demonstrate potential uses for modern machine analytic technologies in the development of geospatial intelligence capabilities.

Earlier this year Lockheed Martin announced they have been contracted to deliver a $20 Million contract to support the AGO in a new program called South COAST. With increasing amounts of information becoming available the AGO is working closely with organisations such as the Australian Space Agency and international partners such as Lockheed Martin to assist with a variety of government missions from humanitarian assistance to enhanced military surveillance.

The first round of funding was announced in March 2021, with New Zealand based companies, Orbica and Xerra along with WA based digital mapping company NGIS all receiving funding. 

Geospatial intelligence (GEOINT) covers relates to the collection, analysis and understanding of the world around us through imagery from aircraft, drones and satellites. In the modern world where the availability of high-resolution satellite imagery, that can be accessed within hours or even minutes, this kind of data is being used more and more.  

This kind of intelligence links geographic data to event data, such as how and why bushfires move across the country, or what is impacting crops or flooding events. It is also being used to support maritime surveillance. It also requires ML to understand it and create context from it due to the large amounts of data collected. 

ML is a part of Artificial Intelligence that utilises a computer's ability to discover patterns and rules directly from the data it is analysing. This process allows for large amounts of data to be analysed quicker than the human brain could. With technological advancement in satellites enabling better access to high-resolution imagery using Synthetic Aperture Radar, there is a particular interest in this area. 

“Detecting vessels in SAR images is challenging because of the complex background, high noise, varying target sizes, and high dynamic range of SAR images. For this project we will leverage recent advances in deep learning to process large-scale SAR images for vessel detection and classification,” said Professor Phung.

The round two funding focussed on a number of “challenge” topics, including, investigating whether ML techniques can identify categories of objects/facilities of interest-based on the input of known dimensions and/or shapes, the impact of masking of imagery by terrain or structures on automated object detection and investigating whether algorithms can be run across commercial SAR satellite data to classify maritime vessels, which is the area UOW have received funding.