'Outback Rover' could help calibrate satellite sensors
Oz robot passes salt lake test
Earth sensing satellites have driven revolutions in farming, mineral exploration and vegetation management, but just like a camera on the ground, they need calibration.
While someone like a professional videographer will check his camera against something like a white balance, earth sensing satellites aren't so conveniently calibrated. The business of satellites like Landsat or TERRA is to take images of the Earth in a range of spectrum bands, and each of the sensors needs to be checked against the known conditions on-the-ground and, if necessary, corrected.
The established process for calibrating an Earth sensing satellite is to pick a location with a known spectral signature, send scientists there with instruments, and perform imaging on the ground at the time the satellite is passing overhead. The two can then be compared and suitable calibration parameters created for analysing the satellite images.
As Dr Alberto Elfes, CSIRO’s science leader for robotics, told Vulture South: “One of the issues with the satellite data is that the more accurate it is, the more useful it is.”
The good news is that Earth has a natural calibration meter. Salt lakes are big (making them easy to identify) and have consistent spectral characteristics, making them an ideal “test pattern” for an Earth sensing satellite. The bad news is that they're often remote, difficult to get to, and hot – like Lake Lefroy, a salt lake around 50 km South-South-East of Kalgoorlie in Western Australia.
CSIRO's "Outback Rover" at the edge of Lake Lefroy. Image: Adam Harper
Dr Elfes again: “Lake Lefroy is big, so it shows up well on satellite images, and it's relatively uniform in colour.” To carry out “vicarious calibration”, he said, scientists travel to the lake, and take their spectral measurements at the surface when the satellite is passing overhead.
“Good spots like Lake Lefroy are often remote and hard to reach, which makes calibration expensive and time consuming”. And an observation run can be spoiled because the scientists made the trip, only to find that conditions are somehow unsuitable for measurements.
Hence the CSIRO's Outback Rover project: it aims to create a vehicle with the same kind of autonomy that lets NASA rovers explore the surface of Mars on their own, but without the multi-billion price tag. The science agency hopes its rover could ultimately be stationed in places like Lake Lefroy, programmed with a mission to take calibration measurements at the right time, and transmit them back to scientists using mobile or satellite networks.
To achieve that autonomy – to operate without needing a scientist at a “home base” operating a joystick – the rover needs smarts other than merely the ability to take the measurements on a schedule, Dr Elfes said. It needs to be able to avoid obstacles without confusing either its location or its mission, and it needs to respond to conditions at the time of its data collection.
For example, he said, if the camera looks up and identifies heavy cloud cover, it needs to be able to postpone its data-collection run. Alternatively, a researcher might identify a colour variation in a satellite image, and request on-ground measurements at a particular location.
It also needs high-accuracy self-localisation, and in the longer term, he hopes to add real-time analysis of the incoming data quality, so that if there's a problem (for example a failed camera) it can switch to a backup unit. ®
Sponsored: Hyper-scale data management