You are not permitted to download, save or email this image. Visit image gallery to purchase the image.
A partnership has been struck between an Alexandra company and Australian geospatial-intelligence organisation to tackle illegal activity at sea.
The $A100,000 ($NZ107,000) deal between the Xerra Earth Observation Institute and Australia’s FrontierSI began this week.
The money was awarded to Xerra to help FrontierSI’s analytics lab programme (AGO Labs) build new industry capability in machine learning and analytics.
Xerra’s work focuses on remote sensing and data analytics, which in the past year culminated in its flagship product Starboard Maritime Intelligence, a platform that uses satellite data and machine learning technologies to monitor maritime activity.
Xerra is one of three successful applicants in the AGO Lab funding.
Xerra data scientist and project leader Joseph Corbett said Xerra and AGO analysts would work together to develop a model to detect anomalies in maritime vessel behaviour at sea, in particular identifying vessels whose behaviour (speed, location, track shape) deviated from the normal activity for vessels of its type.
The work would enable AGO analysts to focus their attention on vessels that were behaving unusually.
This work would be a continuation of research and algorithm development for the Starboard platform, using vessel transponder data and satellite data to analyse vessel behaviour at sea, searching for evidence of illegal, unreported and unregulated fishing, human rights abuses, and other related activities.
Mr Corbett said Xerra was excited to be working with AGO, FrontierSI and their analysts to better understand the questions they asked when looking at maritime vessel behaviour: what "normal" behaviour was, and what signified deviation from that.
"AIS offers a significant potential for gaining maritime domain awareness, but large data volumes and data quality issues prohibit effective manual analysis at scale," he said.
"Our objective is to develop an automated anomaly detection model based on recurrent neural networks — a machine learning technique commonly used to model sequences of data — to learn the behaviours of different vessel types, and ultimately enable us to detect when a behaviour deviates from ‘normal’."
The highlighting of anomalous activity could offer a step change for AGO analysts.
FrontierSI chief executive Graeme Kernich said the programme’s outcomes would improve the way AGO and the Australian Department of Defence worked with companies such as Xerra to test innovations and applications and develop solutions, all with the objective of strengthening partnerships to build geospatial intelligence capability.