AUTOMATIC GLOBAL MONITORING OF THE EARTH FROM SPACE: A BIG LEAP IN SATELLITE IMAGING
Abstract
This year, satellite imaging will be shifting from the photo-interpreter era to automatic monitoring and big data processing. The Eu""ropean Space Agency is launching two earth observation public satellites, Sentinel 2a and 2b, which will yield the first (free) stea""dy flow of multispectral images of the Earth, everywhere, every 3-5 days, at 10 meters resolution.Two US companies, Planet (former"ly Planet Labs) and Terra Bella (formely Skybox) are currently launching fleets of earth observation satellites that will enable to" visit a given site once or twice a day, at a 1 meter resolution for Terra Bella and at a 3-5 meters resolution for Planet, which an""nounces 150 satellites for end 2016. (Other smaller companies (SmallSat, BlackBridge, BlackSky, EarthCube) have announced similar mo"re modest projects.)These major events will completely change the use of satellite imaging. At this time and space resolutions (ho"urs, 3 - 10 meters) one can ``monitor the Earth, namely analyze times series and detect significant change in human activities. Th"us ``Monitoring the ground becomes a reality. Our plan is to make the theory and to experiment an automatic chain analysis that learns a stochastic background model for each site and establishes automatic detection thresholds based on the a contrario theory.Th"is, coupled with the fact that one can acquire high resolution snapshots from private satellites on the same site, will lead to eval"uate the false positive and false negative rate for public medium resolution satellites by using high resolution snapshots as ground" truths. Our goal is to draw change maps for large regions of interest, be it on short time spans and for limited areas, but also fo""r global trends (agriculture, cities, industrial zones, airports, new constructions, natural catastrophes, trafic, crowd motions, wa"r zones).This plan also requires an accurate reconstruction of the Earth to ensure the correct orthorectification of opticalimages. US companies (e.g. Digital Globe) have launched and plan to launch many more high resolution satellites with 3D stereo capability". Their number, which is still insufficient for global coverage, will multiply in the next two years until indeed a high resolution" (< 0.5 meters) coverage of the Earth is possible with frequent < 2 days) revisit time.CMLA at ENS Cachan has developed the first" automatic satellite imaging chain, S2P, which makes automatic 3Dreconstruction from Pleiades or Worldview satellites. We shall als"o construct an automatic 3D monitoring chain.Indeed 3D changes are much easier to detect and analyze than optical changes. Furtherm"ore, satellite images must be ortho-rectified to be comparable, and this is only possible if the digital elevation map is known accu"rately. Change detection on optical images needs an accurate registration which will be provided by 3D accurate ground reconstruction.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 05, 2017
- Source ID
- N000141712552
Entities
People
- Jean-michel Morel
Organizations
- Office of Naval Research
- United States Navy
- École Normale Supérieure