American scientists are working on software that will make satellites smart enough to alert people on the ground to interesting, or potentially dangerous phenomena. The technology could be used to sift through data from Mars, for instance, to identify sites where there are signs of water, or on Earth, to spot flooding in rivers.
The researchers at the University of Arizona (UA), Arizona State University (ASU) and the Jet Propulsion Laboratory (JPL) are developing machine learning and pattern recognition software to teach the satellites to organise the data they send back, so that the most interesting things are returned first.
The software is still very much in development stage, but lab tests have been interesting. The Hydrology group at UA ordered some images from NASA's EO-1 satellite, the satellite they are developing the software for, to see if their software worked.
According to Felipe Ip, a PhD student on the project, it works very nicely: "We didn't know the Diamantina River was flooding, but when we started running the images through our software, it told us, 'Hey, we've got a flood here.' We were delighted because that's just what it's supposed to do."
The flood-detection software compares new and stored pictures of a given region, looking for differences. If things are similar, the satellite does nothing, but if it finds significant difference, it takes more pictures and alerts scientists on the ground.
JPL team members are developing similar software that will be able to detect volcanic activity; and the team at ASU are working to find changes in ice fields.
The three projects are very Earth-focused right now, but the scientists are convinced the techniques will be useful in robotic exploration of the solar system. The idea is that smart space craft visiting other planets can detect and record interesting events, without external input.
The list of potential applications is certainly glamorous. The teams want it to go hunting for volcanic eruptions on Io, cracking ice sheets on Europa, changes in Saturn's rings or the formation of jets on comets.
The flood detection software will be loaded onto the EO-1 satellite for further testing in July. The rest might take a little longer. ®
Sponsored: Ransomware has gone nuclear