Dr. Jonathan Henderson, from User Services Directorate of Information Services at the University of Edinburgh, reports on a seminar he attended recently regarding satellites, big data and intelligence.

Brendan Austin from Bird.i  came to the BCS Edinburgh branch to deliver a seminar 'Satellites, Big Data and Intelligence' as part of the local events series.

Founded by Corentin Guillo in 2014, Bird.i moved to Glasgow in 2016 - joining the burgeoning space industry in Scotland alongside other operators such as Clyde Space and Spire. These 'new space' operators - which have gained prominence with Elon Musk's highly publicised SpaceX ventures - are disrupting the space sector.

Developing 'new space'

Increased commoditisation (e.g. CubeSat using 'off the shelf' components) means it has never been cheaper - nor solely the domain of governments - to launch low cost satellites. Consequently, multiple start-ups have been raising private investment to do just that. Previous barriers are being broken down, with images of geographical areas of interest now being much more economically viable for businesses to obtain, and more easily delivered.

With an increased number of low cost satellite arrays, offering higher temporal resolution of captured images, the hitherto fanciful idea of tracking the world in (near) real time is fast becoming reality. Alongside developments in cloud computing (e.g. AWS) and open source AI some state-of-the-art applications are now beginning to emerge, with disruptors such as Bird.i seeking to extract meaningful insights from the data deluge for their clients.

Satellites are important, and increasingly so with the Satellite Industry Association reporting a 53 per cent increase in the number of satellites launched over the period 2012-2016 than during the previous five year period. The 'new space' in which Bird.i and other start-ups are operating is as small commercial providers - distinct from the hitherto better known 'free to use' providers such as the European Space Agency; and large commercial providers, such as Airbus. Among the commercial providers, a further subdivision into 'incumbent space' and 'new space' exists.

Satellites operating in ‘incumbent space’ are typically large (around 5m in height), complete with a heavy camera payload. Designing these satellites is time-consuming, with high launch costs. As a consequence of fewer large satellites orbiting in an array, revisit rates are less frequent.

The 'new space' satellites are much smaller (around 1m in height), lighter and cheaper. More satellites can be launched into orbiting arrays, with attendant higher revisit rates. The compromise is careful consideration of what sensors are used, given the lower overall possible payload.

‘New space’ operators are seeking to exploit higher revisit rates, acknowledging that for many business needs the typically higher spatial resolution of the larger and heavier sensors aboard satellites in ‘incumbent space’ is not required.

Delivered via an API, the satellite-derived imaging products used by Bird.i can be overlaid on a Google Maps 'basemap'. In many cases, the more recently acquired 'new space' satellite images are used to demonstrate business value to paying customers, for their areas of geographic interest. Time-series data - demonstrating change captured with more recently acquired images - can all be visualised in a web browser.

New intelligence products - business insight from pixels

Bird.i is developing a number of products, including:

1) Oil tracker

The geopolitical importance of oil remains high. The cylindrical tanks in which crude oil is stored at various refineries worldwide can be monitored remotely. The 'roofs' of the tanks 'float', to minimise evaporative losses. The shadow cast by the sides of the tanks acts as a measurable proxy for how full the tanks are, a feature that is measurable by satellites over time. Combining information on the satellite’s camera resolution, and the radius and diameter of the tank, allows a 'fill-in percentage' to be calculated.

Oil reserve estimates are typically published selectively, and monthly. The oil tracker is updated according to image availability, which is more frequent. The higher temporal resolution mitigates against any poor images captured, for example, due to cloud cover (an issue common to satellites with sensors operating in the optical spectrum).

Machine learning - the use of algorithms to assist in the prediction, classification, and clustering of data - is a key technology underpinning the interpretation of imagery that permits analysts of the oil industry to gain insight that would otherwise be impossible.

2) Construction tracker

A feature of the construction tracker is to permit businesses to track the progress of on-going projects. Until now, construction tracking has largely relied on time-consuming, and expensive, on-site monitoring by humans. Accuracy and timely revisits are often compromised. Remote sensing by satellite offers clear benefits.

Initially, pixel-level monitoring of two images - typically taken six months apart - are analysed for each location. Utilising a machine learning model, 'noise' and actual change are separated out, and the output is verified with the customers. A key advantage is economic, with fewer human visits to disparate sites required for monitoring progress.

Future directions

Irrespective of the intelligence product in use, some problems and caveats exist. Creating and validating data sets is time-consuming, and an element of subjectivity exists in image interpretation.

Business use cases vary widely, however Bird.i and other ‘new space’ operators believe there is much as yet untapped potential in exploiting satellite data for environmental benefit - as is increasingly being recognised elsewhere.

About the author

Dr. Jonathan Henderson  works within the User Services Directorate of Information Services at the University of Edinburgh. He has an MSc in Geographical Information Science from the University of Edinburgh’s School of GeoSciences.