SIMVEE - Synthetic Imagery training for Machine Vision in Extreme Environments

"Autonomous cars are frequently in the headlines due to their potential to revolutionise the industry and transform road safety. Similar technologies can be used in the maritime environment however due to the extreme and dynamic conditions more research is needed to address and achieve similar capabilities.

This project will focus on innovate research using a combination of simulated and real world imagery to train deep learning neural networks to undertake object detection classification in extreme maritime environments. We will develop a sensor system for autonomous boats using artificial intelligence (deep learning) techniques to detect objects in extreme environmental conditions. The performance of such techniques is dominated by the volume and quality of training data, but collecting such a set in extreme conditions is prohibitive. We will therefore explore novel ways of combining simulated and recorded data together to develop a system that will detect, track and classify objects within extreme maritime environments.

Currently there are no sensor systems able to detect small objects (such as humans or buoys) in extreme environmental conditions at sea. Hence we will exploit ground breaking research in mixed synthetic and real data training to address this sensing gap. Specifically we will integrate the BMT Rembrandt simulator and ASView control software together to research and develop AI classifier training performance (as well as in verification and validation) in extreme environments. We will additionally integrate the SARIS search and rescue mission planning tool to demonstrate this capability in a fully autonomous real world search-and-rescue scenario.

This project will be led by ASV in collaboration with BMT. ASV is the world leading developer of autonomous surface vehicle systems and has been developing advanced autonomy for these systems for over 3 years. This research has resulted in an advanced autonomy system capable of utilising radar and AIS to complete collision avoidance across a wide range of scenarios at speeds of up to 30knots.

BMT is a leading engineering, science and technology consultancy operating mainly in the maritime industries. With around 1,500 professionals located in 60 offices in Europe, Asia and the Americas we draw upon a wide range of experience and expertise to provide high-quality, high-value products and services.

As USVs are fundamentally limited by sensing ability, full autonomous operation in extreme environments is currently not possible. This project seeks to address this and expand market opportunities."