One of the most common inquiries SEA.AI receives from interested individuals pertains to its capability to detect semi-submerged containers. Detecting a floating container is in most cases straightforward due to its larger size compared to buoys, its rigid rectangular structure, and the temperature differential between the container and the surrounding water.
The essence of SEA.AI's prowess lies in its object detection and identification process. Objects are identified through a technique called 'feature extraction,' which analyzes their color, structure, shape, and temperature (using thermal cameras). Consequently, each recorded object possesses a unique 'visual signature.' However, a single object can exhibit significant variations depending on factors such as the viewing angle, distance, sea conditions, level of submersion, orientation in the water, time of day, weather conditions, sunlight intensity, and its spatial orientation and inclination, among other variables. As a result, to confidently identify an object, SEA.AI often requires input from hundreds of thousands of images.
While passive image collection is valuable, it may not always yield the most suitable data for identifying objects that could potentially cause catastrophic collisions.
In this specific case, SEA.AI's Research and Development (R&D) team embarked from Brest aboard a research vessel (Céladon Association) to gather additional data for their Machine Vision algorithms and to evaluate the AI in authentic navigation scenarios. This crucial step is essential to demonstrate that SEA.AI systems can indeed detect containers and to train the AI for future detections and identifications.
Shaban Almouahed, one of SEA.AI's R&D engineers, elaborates, "We placed a container in the water and captured images from various angles, under different lighting conditions, and throughout the day. We aimed to simulate diverse collision scenarios to assess the system's performance and enrich our database with thousands of images."
What is SEA.AI?
SEA.AI‘s Advanced Assistance Systems provide enhanced situational awareness and surveillance capabilities for recreational yachts, commercial vessels, first responders, and government vessels.
Whether ships, boats, raft, floating obstacles, buoys, and persons overboard: SEA.AI systems reliably detect and identify objects on the water surface, including those that escape conventional systems such as Radar or AIS.
Find out more: www.sea.ai
How does it work?
On-board camera systems autonomously process images and display safety-relevant data on an intuitive user-interface. Intelligent alarming helps crews to make informed decisions.
SEA.AI combines lowlight and thermal cameras with artificial intelligence. Using latest Machine Vision technology, best-in-class deep learning capabilities and an ever-growing proprietary database of millions of annotated marine objects, SEA.AI provides the most comprehensive digital understanding of the vessel’s surface surroundings.
Video realised by polaRYSE - www.polaryse.com
Great initiative, well done. However, it is the container that is floating low down at water level that poses the greatest danger, not the one that is floating this high presenting a similar profile as would another boat. Hopefully the data collected will help the SEA AI to evolve to the necessary level.
Something tells me they may have gone through several levels of submersion. I have no real evidence of that however. They just seemed to really have it together and if they spent the money to do all this, maybe they worked it in many configurations. Who knows. What a great invention !
its all to do with thermal imagining (differences in temperature) so if any part of the container is above the surface then the camera will detect it. even the size of a football. the smaller it is the less range. if no part of the container was above the surface it would not be detected.