We are pleased to announce that we recently completed sea trials on the Royal Navy’s uncrewed submarine. The contract was awarded by the Ministry of Defence to trial innovative sensor technologies that increase platform safety and contribute to extending autonomous voyages to potentially thousands of nautical miles.
Our autonomy-enabling AI engine, FortifAI, was deployed on the nine-meter-long eXtra-Large Unmanned Underwater Vessel (XLUUV), with performance remotely monitored from a nearby support boat. Using the submarine’s inertial navigation system and a combination of periscope optics and an onboard RF receiver, FortifAI passively verified the identity and geographic location of ships during the voyage. Everything was run onboard in real-time using our NVIDIA-based VizEdge device, which simultaneously logged submarine INS data, visual detections, mp4 videos and AIS tracks for post-mission analysis and to allow us to improve and add new capabilities for the future.
The trial presented an excellent opportunity to test FortifAI’s new ability to conduct semi-unsupervised self-training. The “Dataset Pruning” process is equivalent to a teammate having a first pass at checking the quality of detections before final human verification. We found that it significantly improved the speed and overall accuracy of the data preparation phase, especially when highlighting the presence of distant objects that a human operator may have overlooked.
In addition to improving the platforms’ real-time situational awareness, FortifAI compiled a list of the last known positions of so-called “dark vessels”, identified from satellite imagery captured moments before the voyage began. The relative distance and bearing of the unknown vessels, along with ship length derived from the bird’s-eye view from space, were stored on the device and graphically indicated on the live video as augmented reality markers.
We look forward to expanding on our research in semi-unsupervised learning and advanced computer vision for maritime intelligence with a leading manufacturer of unmanned surface vehicles in 2022.
About VIZGARD Ltd
VIZGARD is an award-winning computer vision start-up on a mission to build a safer future by automating mundane 24/7 security system monitoring. Optimised for high-performance on-the-edge, their autonomy-enabling AI engine “FortifAI” provides long-range visual analytics, external sensor fusion and camera platform control to reduce false alarms and operator burden.