the example of disaster management - gipsa-labjocelyn.chanussot/teaching/sicom_1_2.pdf · grenoble,...

Grenoble, 2008/09/08

VHR Systems Missions Challenges Development Image Processing New missions

The example of disaster management









New constraints

I Short system time responseI Much more than a short revisit cycleI Reactive satellite taskingI Quick image availability

I Actually, information availabilityI High spatial resolution + wide swath

I Being able to see details ...I ... without actually knowing where to look for them!

I High spatial + spectral resolutionI Geometric and radiometric information to finely discriminate

objects and their characteristicsI Permanent (geostationary) and high resolution (low orbit)

observation



How to find a solution

I Don’t worry, space agencies have clever people to solvethis kind of problems

I Inter-satellite links for retasking and up/downlinksI Agile platforms for one-pass image mosaicsI Virtual huge telescopes by aperture synthesisI Deployable antennasI Constellations of smaller, cheaper satellitesI Many clever devices that you thought couldn’t exist out of

Star TrekI And other top secret things I can’t talk about ...


VHR Systems Missions Challenges

Still any challenge for IP people?

I The solutions cited above are still expensiveI An IP-based solution is softwareI Software is cheap. Best software is free!

I Users’ expectations go further and further (and quicker)I Everybody has a GPS receiver, Google Earth, Open Street

Maps, etc.I Thus new imaginative uses for geospatial information

appear every dayI Updating software is quicker and cheaper than building a

satellite



New value for old sensorsI RS scientists are good at finding new uses for existing

sensorsI some examples were given beforeI sometimes is just a proof of concept: DEM extraction with

SPOT 5 P+XSI sometimes operational applications are built upon these

ideas: Persistent Scatterers InterferometryI In some sense, the International Charter Space and Major

Disasters uses this approachI Most of the available systems are not designed for disaster

managementI 2 paths leading to these new ideas

I a real need triggers the idea:I how can I deliver this fancy information with this set of old

crappy sensors?I take the system to its limits: squeeze the resolution,

interpolate, deconvolve



Existing ancillary data

I Why extracting from images what you already (nearly)have elsewhere?

I Maps, vector data bases, the Web, etc.I Remember Kalman

I Update your guesses (ancillary) with new data (RS)I However

I How to assess the quality of ancillary data?I How to compare ancillary data and RS images?

I Level of representation?I How to assess the quality of the updated geoinformation?

I Measuring the performances of the algorithms: reliability,precision, accuracy, etc.



Measuring the performances of a system

I AssessI the quality of the input informationI the performances of each individual processing blockI the quality improvement introduced by fusion

I So the quality of the output is known



Fusing new kinds of data

I Beware of fancy, useless fusionI Have you heard of optical/SAR fusion? Me too.I Have you seen real applications of it? Me neither!

I The way to put it is:I Given a user need, which is the most straightforward way

(data availability, price, delays) to get the information?I If this means that fusion is needed, go for it.I Usually, pixel-level fusion is not the best way to do it

I except for pan-sharpening, and maybe medium resolutionmulti-temporal series

I On the other hand:I Fusing any kind of data may be possible. Just choose the

appropriate representation level.



New computing approaches

I VHR means huge data volumeI Parallelism, streaming:

I These are techniques allowing to overcome computationtime and memory capability limitations

I They need to split the data in order to process themI Algorithms which can not process data by chunks are dead!

I Scalability of algorithms is crucial for end-user applications

the example of disaster management - gipsa-labjocelyn.chanussot/teaching/sicom_1_2.pdf · grenoble,...

Documents