​PSMA Australia launches Geoscape location analytics tool

Geoscape uses a combination of satellite imagery, crowdsourcing, and machine learning to capture attributes of Australia's built environment and link it to a geocoded address.
Written by Asha Barbaschow, Contributor

PSMA Australia has officially launched its location intelligence tool Geoscape, which combines satellite imagery, crowdsourcing, and machine learning to capture attributes of the built environment and link each building to a geocoded address.

Geoscape is a collection of information across Australia's 7.6 million square kilometres that represents more than 20 million buildings, vegetation, the surface cover of the ground, height information, and attributes associated with buildings, which is captured in a relational database.

The government-owned provider of location datasets opened its Adelaide, South Australia, and Canberra, Australian Capital Territory datasets late last year, but PSMA held onto its official launch to coincide with the release of its Sydney dataset.

The data includes building footprints and heights, roof construction, land cover, tree heights, the presence of solar panel installations, and swimming pools, and is visualised using mapping platforms and 3D modelling tools, such as the Australian government's National Map initiative, which enables the visualisation of any publicly available spatial dataset.

PSMA CEO Dan Paull expects that organisations and government entities will analyse the available data to gain new insights into the built environment and improve the way they plan and operate.

The Geoscape project kicked off when PSMA identified the need in the market for a particular collection of data that the organisation could see was going to have wide-reaching impact in terms of improving quality of decisions and the analytics that could be performed.

"There doesn't currently exist a comprehensive data product of this type in Australia; we actually don't have an analytics product that enables us to understand where all the buildings are in the country," Paull told ZDNet.

"Given the amount of money we've all invested in those buildings right across the country, being able to understand how to manage those assets as a collection across the whole of the continent is actually difficult to do without core information -- the specific details of those buildings."

Paull explained that PSMA then had to work out how to meet the specification, and consulted with vendors about the limitations existing technology has, along with how the organisation could collect the data. What resulted was a partnership with US-based commercial space imagery and geospatial content vendor DigitalGlobe.

DigitalGlobe has five high-resolution satellites zipping around the earth at about 7.5 kilometres per second and collecting imagery down to a 30-centimetre resolution.

"One of the things that's unprecedented about what we've done here with this partnership is that any one of those technologies on its own has been heralded as a disrupter and game changer," Paull explained. "What we've done is connected them all together and then applied them at an extraordinary scale."

In developing Geoscape, PSMA and DigitalGlobe had to collect current inventory using short-wave infrared, which allows for intense material identification, such as determining what roof tiles are made of.

The duo then used crowdsourced materials to train machine-learning models used to create the dataset. To achieve this at the scale required to capture information on an entire continent, PSMA and DigitalGlobe partnered with Amazon Web Services.

What is achievable with the technology, Paull said, is "mind boggling".

"I think one of the biggest challenges that we've found in taking this to market is actually getting past the initial hurdle of disbelief," he said.

"But it really is possible, and the quality of the content is really quite remarkable."

In developing the product, Paull said the potential for the capability to be deployed into developing countries to support the rollout of information that's necessary for economic development was realised.

"We've developed quite a powerful process for capturing very valuable information quickly and relatively inexpensively, and those same techniques can be used to help accelerate programs for information capture in developing countries," Paull explained.

"We've had a lot of interest from the World Bank and from the United Nations around this type of capability, supporting initiatives and programs that they're pursuing including the United Nations sustainable development goals."

With analyst firm Gartner predicting 20.4 billion Internet of Things devices to be deployed by 2020, Geoscape is also expected to play a large part in its success by linking a wide range of information generated about the lifecycle of the built environment.

Paull said the rollout of datasets will continue across this calendar year, and conclude in early 2018.

Editorial standards