Positive experiences with monitoring movements on Earth via satellites have led to a new research collaboration between DTU and a number of private companies and defence institutions in Scandinavia. The project—which is called Bifrost—focuses on developing smart satellites which can analyse image data using artificial intelligence already in the satellite and thus lead to less expensive and faster systems.
“Earth observation from space is a rapidly growing area in these years. We’re experiencing great interest in this field, including for use in the area of defence and security, where surveillance and territorial assertion can be significantly strengthened with the new technologies we’re developing. Systems of this type offer a great potential for Denmark in a huge area like the Arctic,” says Head of DTU Space Henning Skriver.
Today, the observations are made by retrieving large volumes of optical images and radar images from satellites and sending them down to Earth, where they are then processed and sorted. The images can be used for a number of purposes, including to detect illegal ships, territorial violations, drifting icebergs, and forest fires. By performing the data processing using artificial intelligence in the form of machine learning and deep learning algorithms in the satellite, the process is made more efficient.
“By using artificial intelligence in the image processing in the graphic processor—the GPU—in the actual satellite, data can be processed quickly in space, after which only the most important data are sent down to Earth. This makes the process both faster and less costly because much less power is used, and the satellites can therefore be smaller,” explains Henning Heiselberg, Head of Security DTU, which is an interdisciplinary security and defence technology research centre.
The satellite discards unimportant data
The artificial intelligence is trained to find interesting objects in a given area and send data about them back to Earth. Here they can then be analysed further. For example to determine whether this concerns a ship sailing illegally in an area, or a drifting iceberg.
“We’ve tested and trained the system on ships, icebergs, and forest fires on land. And we’ve had good results, which we’re now building on,” says Henning Heiselberg.
In the new project, data is used in the form of optical images and multispectral recordings in the infrared region recorded from space. They are combined with data from—for example—radio antennas on Earth. And here artificial intelligence (AI) works effectively when unimportant data has to be discarded so that only the necessary information remains. Data from radars in space and on Earth can also be used.
The new project adds a layer to DTU’s current research and development within space-based defence and security technology.
Illegal ships can be identified
The development of the new satellites builds on several years of research under the Dark Ships project, the purpose of which was to develop a satellite-based system to search for illegal or distressed ships in the Arctic. Surveillance of ships in the Arctic is increasingly relevant, because more areas here are becoming navigable as a result of global warming, thus resulting in increased ship traffic.
The Dark Ships project was a collaboration between DTU, the company Gatehouse, the Danish Armed Forces and Joint Arctic Command. The purpose of the project was to develop satellite surveillance of ships sailing under the radar by switching off the so-called AIS transponder. When the transponder is switched on, it emits data on the ship, as well as its route and position, etc. All large ships must have such a transponder according to international maritime regulations.
It will typically be switched off on ships that violate Danish territory and international legislation. This may be illegal fishermen, smugglers, military vessels, and ships that dump oil or are carrying cargo containing weapons or other goods that are subject to sanctions.
Dark Ships are located by using both optical images and radar images taken by, for example, ESA's Sentinel satellites or other satellites with antennas that can pick up radio signals from Earth.
“If a ship hides from one sensor, but we capture it on another, it will typically be a suspicious vessel that should be checked out,” says Henning Heiselberg.
Database with 10,000 ships
Artificial intelligence (AI) developed at DTU analyses and compares the large data volumes from these sources and finds patterns that could—for example—be an illegal ship. In this way, the system can relatively quickly locate a drifting iceberg or a ship with a switched-off AIS signal and distinguish the two from each other. This reduces the number of false alarms, which can save—for example—the Joint Arctic Command precious time and resources.
“These are key technologies that can make territorial assertion and surveillance over long distances much more effective. Similar systems are used to map sea ice to achieve safer ship routes in the Arctic and in connection with rescue operations,” says Henning Heiselberg.
The researchers now have a database of over 10,000 known ships and even more icebergs.
“We’ve shown that the concept works. We’ve developed algorithms based on artificial intelligence that can themselves analyse the large volumes of data and distinguish between icebergs and ships. And there is potential to discover and recognize many other interesting things with these types of systems,” ascertains the DTU researcher.
The researchers have also discovered that radars on ships and on the ground can be found via satellite data. In fact, it turns out that noise—interference—can form between the satellite and radar signals. Based on the interference pattern, the position and type of the radar can be calculated.
Collaboration with industry
DTU Space has long expertise in space-based technology for monitoring the Earth. This knowledge is now increasingly also being translated into defence and security technology.
“We conduct research into and develop technologies, provide scientific advice, and educate engineers with the right competences in the field. And we use all our knowledge to develop security technology in these new projects,” says Henning Heiselberg.
“Our activities fit perfectly with the growing interest in strengthening Denmark’s defence. In collaboration with industry, we can contribute with a number of practical tools for this purpose.”
