Detect and Classify Mine-Like Objects
Automatic Target Recognition (ATR) is the tool of choice for analysing side scan data from unmanned maritime systems. Designed as an assist tool, the ATR provides a post mission analysis (PMA) workflow with robust, reliable results, regardless of the data volume. We have worked with the US, UK, Dutch, Belgian, Australian and New Zealand Navies to provide them with our ATR system.
Analysing large volumes of data over long periods of time results in fatigue and can impact on performance. Our ATR uses fast, machine learning techniques to detect and classify mine-like objects from the side scan sonar data. It provides a measure of how ‘mine-like’ each of the targets is and can be used to decide if the target is a mine or a false alarm. An updated workflow for common tasks, such as managing multiple views and fusing duplicate contacts, improves mission tempo.
Our ATR System provides consistent, repeatable PMA performance for improved mission tempo.
The Mission Review Tool automatically provides multiple views of the contacts. This enables you to see all observations from different passes instantly without the need to mark them individually.
The fusion module enables you to merge the detections generated by the ATR algorithms. This removes the need to manually review the detections, to combine, or remove duplicates, and alternative views.
The complexity of the different sea-floor terrains is analysed and presented using a geo-referenced colour map. This can be important in assessing whether further PMA is appropriate, and how to conduct this PMA.
The ATR System is built on top of our commercially controlled open architecture. A Software Development Kit (SDK) allows integration of third-party ATR algorithms or Fusion algorithms. World-wide government and commercial customers have already benefited from this capability.
All of our products are designed to help manage unmanned maritime systems, ultimately providing situational awareness across all assets within the battlespace. Our additional modules provide specialist analysis, processing and training.
Advanced machine learning algorithms require the provision of example data from a relevant operational environment with good examples of target types. In order to allow customers greater flexibility, the ATR Tuning Tool can be used to update or retrain the ATR algorithm to work in new environments or with new target types.
Embedded ATR uses fast machine learning techniques to run real-time. Neptune logs all the contacts, including an image snippet, and can also optionally add dynamic reacquire tasks for any vehicle in the squad. The results are fully compatible with the PMA ATR System, and can be loaded directly upon mission completion.