End-to-end technology stack translating photons into information

Our approach to timely, sub-meter imagery lies at the convergence of Silicon Valley ingenuity and world-class aerospace expertise. While we are building, designing, and operating our own fleet of high-performance imaging satellites, it's really all about the data platform.

Analytics Platform

Big data. From space.

Imaging satellites generate vast amounts of data, and require an Internet-scale data approach to unlock their full potential.

It's the 21st century. Instead of using traditional relational database management systems, we use open source technologies and are pioneering the use of Apache Hadoop for tasking, storing, and analyzing large quantities of satellite imagery and derived data. 

We have developed a custom framework, called BusBoy, capable of running scientific computing algorithms written in C or C++ as Java tasks within Hadoop. This framework enables rapid production deployment of future algorithms against Skybox data, and is already used extensively by our team to develop our own advanced satellite imagery processing tools. 

While we believe Skybox-collected data is critical to informed decision-making, it is only one part of a mosaic of complicated data sets. It is a crucial data piece in the larger puzzle of understanding the world. We all win when more data plays together — so we designed our data platform to be completely sensor agnostic.

UAV builders, satellite operators, social aggregators: get in touch with us to explore the benefits of including your data and algorithms within our platform. 

Imaging Chain

Photons to pixels. Unprecedented flexibility.

We have developed cutting edge technologies across the complete imaging chain. From capturing photons in orbit to processing pixels on the ground, we have the flexibility to provide global businesses unprecedented transparency.
The laws of physics make high-resolution imaging very challenging from small satellites. Our researchers and engineers have experience working on some of the world's most advanced imaging technologies, allowing us to approach the theoretical limits of performance.

Taking quality pictures from space is hard. Imagine a camera more than 600 kilometers away from its target, moving at over 7 kilometers per second, trying to see an object less than 1 meter in size. At this distance, the law of diffraction limits the smallest ground object that can be imaged through a telescope based on the diameter of that telescope. Using sophisticated, proprietary manufacturing techniques, our telescopes are able to approach the limits dictated by the laws of physics – enabling higher quality imagery than ever seen before in similarly sized optics.

Cameras and Sensors
Traditional satellites capture imagery like a copier machine — with a line scanner that builds up images row-by-row. These sensor arrays capture lots of light, meaning less of the grainy noise you see in a cell phone picture taken in too dark of a room, but they make satellites extraordinarily complex. Skybox uses the latest generation of two-dimensional imaging sensors, coupled with sophisticated, proprietary camera electronics to dramatically simplify our systems in space. 

Contrary to line scanners, these sensors allow us to capture a video stream. The many individual frames of the video can then be combined in order to provide very high quality, feature-rich images that allow users to see vehicles and other similarly-sized objects that drive our global economy. We are also pushing the boundaries of on-orbit processing, using advanced on-camera image processing algorithms to enable temperature-adaptive correction of non-uniformities, before the pictures are ever downloaded. 
Each of our satellites generates over a terabyte of data per day. This data has to be processed, stored, and indexed in a way that is flexible, affordable, and scalable. At Skybox, we are pioneering the use of Hadoop for satellite imagery and other geospatial applications. We have built API's that bridge the gap between Hadoop and image scientists.

Our image science team has used this technology to develop PixelLoom, an advanced suite of pixel processing algorithms including frame-to-frame registration, super-resolution, non-linear denoising, pan-sharpening, color enhancement, and a host of other features. All of this is built against Hadoop, which means instant scalability and provides us the ability to continuously improve our algorithms and the data we are able to provide you.

And the best part? None of our software is Skybox-specific. We'd love to talk about processing your data in our infrastructure, whether it's from an un-manned aerial vehicle, satellite, or ground-based video camera.

High-Performance Satellites

Small satellites. Huge performance.

Skybox was founded out of the CubeSat community and we are ardent believers in the power of commodity, commercial electronics to change the cost of doing business in space. Our industry is experiencing tremendous disruption through reductions in launch costs and modern commercial electronic components that are capable of high performance and high reliability in orbit.

Traditional satellites capable of taking imagery at better than 1 meter resolution weigh thousands of kilograms, which makes it prohibitively expensive to launch enough of them to capture timely imagery. We have produced similar performance in a box 20x smaller by breaking open the many black boxes that define traditional systems and creating an optimized design using automotive grade electronics. 

The circuitry that drives our satellites - providing power, attitude control, communications, thermal management, and imaging support - are about the size of a phone book and consume less power than a 100w light bulb. We've integrated the latest, greatest, and fastest commercially available FPGAs, processors, and memory to ensure our small satellites pack the largest possible punch.

Their small size means we can afford to launch lots of satellites, and provide you lots of timely, sub-meter imagery and video, along with powerful derived analytics.