Airbus: Charting new territory with Platform

The Intelligence business line of Airbus Defence and Space uses to build a scalable, online platform, enabling customers to access petabytes of satellite imagery in real time.

Adds millions GOOGLE CLOUD RESULTS of square • Reduces the time required for customers to access satellite kilometers imagery from hours to less than half a second of imagery per day • Delivers a scalable, cloud-based platform for satellite images with no limits on size or users • Provides access to satellite imagery for smaller organizations with streaming technology, , and innovative pricing schemes

Airbus Intelligence, a unit of the Airbus entities that benefit from high quality we wanted to move forward, Defence and Space division, delivers imagery such as agricultural companies If we needed to have our comprehensive, up-to-date satellite and oil and gas firms. To stay ahead satellite images online and imagery and data to more than 1,200 of the curve, Airbus Intelligence was provide our customers with customers in over 100 countries around keen to build on its success and real-time access. Google Cloud the world. Since its founding more than redesign its satellite imagery platform Platform helps us to do that while three decades ago, Airbus Intelligence for scale and speed. To do that, it maintaining our capacity to offer has expanded its roster of clients from turned to Google Cloud Platform (GCP). very large amounts of imagery governmental security to include other and data.

Laurent Gabet, Optical R&D Manager, Airbus CIVIL

DEFENCE

AGRI SECURITY CULTURE

LAND INTELLI ADMIN. GENCE

LOCATION

AVIATION

OIL & GAS FORESTRY MARITIME Google Cloud Platform for security and scalability

roducing satellite imagery is a P complex business. Airbus currently operates and has access to 14 different satellites with the ability to take images all over the world every day. Traditionally, once the satellites had captured the images, Airbus would process them and download them into the company’s library. From here, customers would choose the images they wanted from a catalogue and then several hours later, they would receive them.

Over time, as technology grew more sophisticated and Airbus added to its fleet of satellites, it built the capacity to provide better quality data and imagery at increasingly large volumes. However, its existing solution could not speed up the production and delivery of its image and data products to meet the growing expectations of its customers.

Airbus decided to create a new platform, called OneAtlas, that would provide customers with the images and data they needed as quickly as possible without compromising on quality or the range of coverage. In addition, the specialized type of imagery that Airbus provided meant that customers had to have the right technical capability and capacity to receive and use it. With the found the managed services to be very efficient and new platform, Airbus could innovate with effective for delivering high-level information to our different image types and solutions to “We enable smaller organizations to use the customers. Google Kubernetes Engine, in particular, is really the company’s products. “We really wanted heart of our system now.” to democratize access to our satellite imagery,” says Jean-Michel Darroy, VP Laurent Gabet, Optical R&D Manager, Airbus Head of Services Growth at Airbus.

Airbus trialed seven separate cloud- which allowed customers to stream to extract information from the images, based options for the OneAtlas Platform. images of the whole world instantly the company could scale its compute It chose Google Cloud Platform as or download them at a range of power up and down as demand the winning solution, largely because different resolutions depending on their increased and decreased. Cloud Pub/ of its facility with deep learning and needs. This demanded extensive data Sub was used for messaging and the image processing, the Google team’s capturing. In addition to its own well- data was held in Cloud Datastore, for customer service, and robust security honed image-processing techniques, easy scalability. models. “Google has an impressive Airbus trained deep learning algorithms reputation for security and that was in TensorFlow to recognize cloud cover. With the images processed, Airbus key for us,” says Laurent. As a key After switching to TensorFlow, Airbus extracted the data and placed it in the component of OneAtlas, found that its error rate for detecting existing library. From here, the customer holds hundreds of terabytes of data and clouds dropped from 11% to 3%. could access the imagery as part of a imagery that’s easily accessible in the seamless, connected digital atlas for cloud and doesn’t require provisioning To process the images, Airbus relied streaming or downloading. Airbus also and maintaining large banks of servers. heavily on Google’s managed services. wanted analytics to provide high-end The first component of OneAtlas With Google Kubernetes Engine services to complete the offer for the released was the Basemap product, orchestrating the various methods used benefit of the customer. Focusing technologies, broadening horizons

ith Google Cloud Platform, Airbus Other services will include industry- Whas built a highly secure, cloud- oogle Cloud Platform specific analytics services for Agriculture, based platform, capable of delivering Maritime, and other markets. In addition, gives us access data and imagery to its customers at G it will add a Software image- unprecedented speed. Customers can to much more compute processing tool, a data-led product built now stream images with less than a from analytics running on satellite and power than we had in the half-second delay, compared to a wait drone imagery and, finally, a Platform of hours or even days with the previous past, so we can do so as a Service offering to allow its less system. The robust, global network of specialized customers access to some much more than we used Google helps ensure that the OneAtlas of the company’s own tools and even platform can deliver the same high- to. Back then, we worked some third-party analytics. quality service to its customers wherever on projects. Now the whole they are in the world. As well as perfecting its current offerings, world is our playground. the Airbus Intelligence has also been Meanwhile, Airbus has been able to take looking to the future, exploring new tools advantage of a highly scalable storage Laurent Gabet, Optical R&D such as G Suite, to optimize its internal solution, adding millions of square Manager, Airbus communications, and TensorFlow kilometers of imagery every week. This to evolve new techniques in remote currently amounts to 500TB of data per of customers now who might not have sensing with deep-learning models. year, but with the planned addition of been able to buy satellite imagery in the According to Laurent, this kind of data more satellites and high-altitude drones, past,” says Jean-Michel. For Laurent, as extraction, designed to identify objects the company expects to add 2PB impressive as the technical benefits are, such as land and sea vehicles, would annually. it is the scope of Airbus’ ambitions that have been more difficult with traditional have been most greatly affected. methods. The new platform also enables Airbus to make its image products much more Thanks to the scope and flexibility “Back when we were still deciding affordable for smaller organizations, of Google Cloud Platform, Airbus on which solution to use, we felt that such as NGOs or farmers who can take Intelligence is hard at work on a range Google had the clearest roadmap for advantage of newer, lower-cost options of additions to OneAtlas which include this kind of technology,” says Laurent. like streaming data for a limited time or expanding its Data as a Service offers “Deep learning has been a real problem only paying for the images they use, beyond the Basemap service to include solver for us, and Google Cloud Platform instead of more traditional subscriptions more high-resolution imagery, elevation provided just the right solution to help models. “We have access to new types data, and automated change detection. meet our market needs.”

Airbus Defence and Space Australia, Brazil, China, Finland, France, Germany, Hungary, Singapore, Spain, United Kingdom, United States © Airbus DS 2013 @AirbusDefence www.intelligence-airbusds.com