Boffins crowdsource hunt for 'Planet 9'
AI isn't ready to find planet beyond Kuiper Belt. So you - yes you - have the chance to scour thousands of images
The Australian National University (ANU) is recruiting citizens to look at hundreds of thousands of images, in case they can find the mooted-but-not-yet-discovered “Planet 9”.
The long-dismissed idea that there's an undiscovered planet beyond Pluto was revived in January 2016 when CalTech boffins claimed they'd spotted perturbations in Kuiper Belt objects they attributed to it.
Their first key estimates were that the object was about 10 times Earth's mass, with an eccentric orbit that requires between 10,000 and 20,000 years to orbit Sol.
In March 2016, the researchers (Konstantin Batygin and Mike Brown) had some vindication when the SETI Institute's Michelle Bannister revealed a previously-undocumented Kuiper Belt Object that partially matched their predictions.
The Australian National University (ANU) group led by astrophysicist Brad Tucker explains if Planet 9 is out there, it's certain to be in observations somewhere - we just haven't spotted them yet.
Hence their crowd science initiative: they've published hundreds of thousands of images from the ANU SkyMapper telescope at Siding Spring Observatory.
Once launched, the SkyMapper images will be hosted at Backyard Worlds, here. That site is separately running a “Planet 9 search” using images from NASA's WISE mission.
The Register asked ANU astronomer Brad Tucker why eyeballs are needed to spot objects, in a world full of artificial intelligence and relatively cheap processor cycles.
Part of it is passion, he said: enthusiasts are very good at this kind of work.
Why no AI?
The Register asked Tucker to flesh out why the search isn't using artificial intelligence.
Sometimes, he observed, the citizen science angle of a project is to get people to help train the computer. In the case of “Planet 9”, however, it's a different problem.
“There are so many possible ways this object could be moving”, he said – its orbital distance, orbital inclination, and speed – “so you're not just looking for one characteristic.”
To apply AI to the problem, “we'd have to have training sets on a huge range of parameters – or, you can say 'get the public involved, and they'll be more effective, so why not?'”
To make things easy for searchers, the original SkyMapper images have been processed to apply a colour-code to objects moving between one image and the next (on the same patch of sky).
Since everything else in an image – the stars and galaxies – is simple black and white, “anything that moves just pops out at you.”
That presentation, Tucker added, makes things a lot easier for the citizen scientist than trying to comb through the relatively poor resolution NeoWISE images.
And it won't only be “Planet 9” the search will find, he said. There will be “dwarf planets, thousands of asteroids, maybe even a comet or something else”.
Alas, if you discover Planet 9, you'll have to find someone else to name it after, because the International Astronomical Union doesn't let you name things after yourself. ®