I previously wrote about the Geospatial 2.0 workshop held earlier this year on the Space Campus in Harwell.
http://www.acgeospatial.co.uk/blog/geoworkshop/
Ever since this workshop I have started seeing more articles and I have become much more aware of the ‘commercialisation of space’. In April this year The Economist wrote a leader on Space Exploration.
machine-learning can mine (EO data) for information on crops, shipping, traffic, wildlife or the environment https://t.co/sAFlPgXUB2
— Andrew Cutts (@map_andrew) August 26, 2016
“The technological progress that has put supercomputers into the pockets of half the world has made it possible to do a lot more in orbit with much smaller spacecraft. A generation of entrepreneurs forged in Silicon Valley—and backed by some of its venture capitalists—are launching highly capable new devices ranging in size from shoe boxes to fridges and flying them in constellations of dozens or hundreds. Such machines are vastly more capable, kilo for kilo, than their predecessors and cheaper, to boot. They are making space interesting again.” http://www.economist.com/news/leaders/21705825-new-discoveries-intelligent-devices-and-irrepressible-dreamers-are-once-again-making-space
There certainly seems to be a trend toward ‘Earth Observation 2.0’ (a move towards more automated processing and faster information analytics). In my post about 3 reasons to revisit EO, number 3 was that there is an ever increasing amount of satellite launches.
http://www.acgeospatial.co.uk/blog/3-reasons-eo-oil-and-gas/
Machine Learning
Machine learning / deep learning on EO data is becoming increasingly relevant with the ever expanding volumes of data. In fact, this article suggests that “Earth Observation is the most crowded segment among satellite-related startups”http://www.huffingtonpost.com/entry/spacetech-ecosystem_us_5793d5cbe4b0b3e2427c6942
Articles are asking if Earth Observation data is a multibillion dollar opportunity or a dud “Some companies… are doing things like predicting retailer profits by counting the number of cars in store parking lots, and monitoring construction and manufacturing rates in China. This is only the tip of the iceberg, and the full multibillion-dollar potential of geospatial big data is only just being unlocked.”
There is a really good talk given by Boris Babenko of Orbital Insight (just one of the companies mentioned in the above articles) that runs through an introduction to EO all the way through to the amazing measurements and analytics they are making with deep learning.
“Rockets are getting cheaper, technology is getting better and looser regulation. 2014 rules were relaxed. Traditionally satellites were big, and very expensive to launch. Because they were so big and expensive to launch into space, it means there is only a handful of them. This means revisit time is not ideal. However, the cube sat is changing this, much smaller, cheaper, many can be sent up. This means the revisit issue is less of a problem. However, the lens is smaller so resolution is limited – but still good.”
More data than ever before
This week a collaboration with DigitalGlobe, CosmiQ Works and NVIDIA launched SpaceNet on Amazon Web Services.
💥new training data for machine learning@DigitalGlobe & @awscloud drop:
– 2k km2 50cm imagery
– 220K bldg footprinthttps://t.co/pfFatL2qEb— developmentseed (@developmentseed) August 25, 2016
“The current SpaceNet corpus includes approximately 1,900 square kilometers full-resolution 50 cm imagery collected from DigitalGlobe’s WorldView-2 commercial satellite and includes 8-band multispectral data. The dataset also includes 220,594 building footprints derived from this imagery which can be used as training data for machine learning. This dataset is being made public to advance the development of algorithms to automatically extract geometric features such as roads, building footprints, and points of interest using satellite imagery.” https://aws.amazon.com/public-data-sets/spacenet/
Why now?
https://soundcloud.com/cbinsights/cb-insights-podcast-i-jimi-crawford-orbital-insight
Three reasons
- We have never had so much data so readily available to us and the cost of this data is decreasing.
- Fundamental change from running on GPUs instead of CPUs
- Evolution of deep learning and neural networks on image recognition and classification.
As the Econmist says “They are making space interesting again”
Want to learn more about GIS and EO for Oil and Gas? Then my page contains all my blogs, plus case studies and links http://gis.acgeospatial.co.uk/