Add to Watchlist

ORFEO ToolBox status report


Citation of segment
Embed Code
Purchasing a DVD Cite video

Automated Media Analysis

Recognized Entities
Speech transcript
this term as it instances seems that the system is below within
the dualistic to be generally assume the goodness of the cylinder so please with manually is in the homes of the budget have thank you you will stay in space OK uh I work at the French face urgency and was to also speak about more deeper image processing and why working on your filter library which is an open-source library for image processing it was that the title of my presentation went the statute recall that I that I I know that it's a little bit of to be prone to little bit more or less famous in the GII was westward my presentation will be 1st introduction to intervene in the 2nd part is about this is about the new features in the last in the last 3 years of the library as will press keep a little bit uh at that point the solution to to to to get the new features so things to know about
what the but it is an image processing library of the Centre for Remote Sensing is free and open-source was developed that 30 years ago 10 years ago we are still children uh it was founded in developed by the French space agency which was developing at that time
I resolution optical satellite which call clay had operated by elements and they wanted also to develop tools to relates to to to to to make possible for users to get useful information from the of the satellites of standardized costing millions of euros and then put some money left less money to develop a tool and software to to other people and its use that can answer but it was the new quite generic way it's not all it's no use at the European Space Agency opportunities for instance use as 1 of the library in the ground segment for process until 2 images and also there are lots of expectations to to rule out to be will be used now at the school's mission exploitation platform is a little bit what markers was presenting the fact that we want to put users algorithms near the data uh and there are few projects is act is educationally where well attendees using also is suppressed press it's using and build on the shoulder of their of giants of course g down everybody when using up but also always use software for and some modeling and lots of dedicated to feature extraction image processing software like opens its the data are capable and other things too the build up of this processing and they do not have to take time at the end of the by streaming the data and using the trading OK it's not the not always your software we have seen the that incubation phase we are listed yesterday as unresponsive in the table I'm not really agree with this but again advent been a presentation on whether on on this that is really it's walking along progress that we went far from the uh so as was doing open source as I said this was developing a satellite and 1 that the the image from the satellite is use of as many users as possible so making an open source toolbox to be able to analyze the data it was it seems like an overview decision at that time it was not not so not so obvious but still the beginning there they wanted to to make it a political function algorithm so it's better reading completely stop of functions well really targeting uh image processing and being able to from the raw data to extract useful information at the beginning because we oriented on high-resolution optical data which is quite an also in the world are remote sensing I would say that the the the the way that it was done in the the genericity of the library and also of image processing algorithms and I pursue to use to use it in different contexts of all different applications so proper preprocessing of course the atmospheric elections as as much as well uh and it is this type of algorithm ot electrification rectification of course the major electoral manipulation that ought to be used as a key using what's what is the available in of these and those the out of feature extraction and information extraction algorithms edge detection keyboards computing local statistics for the metric indexes and then with this generality what you want to do is look at classification and segmentation excetera oxygen so that that's really the target of of quality with human segmentation algorithm the mean shift that was presented in the 1st presentation is available in what you for for a long time now because it's really uh working well expression in high resolution images uh welding also of object-based image analysis and we've got also although they're not of classification and unsupervised also provides there also on existing specialize library like the busy and and also opens the the the the bridge between the TVs to be able to access full dimensional learning algorithms available in open cv directly in what OK so
it's something the example the change detection algorithms
but want to do so we want to be is this algorithm is called a diet that direction detector it's quite famous user not indeed Geoscience uh community you go to I just conference you will see a ton of Postel's using this is also fully automatic it again what some down and swapping pretty well but it's not easily available open source software open source library can encodes that that this kind of algorithm that the 2 of the idea in order to be to to make it possible for users to have access efficiently of that type of algorithm is also 1 is the
line that they cut out going the energy line segment detector is also quite famous but more in the pattern-recognition the analyzes word I would say that it's also fully automatic free really fast and working pretty well as a preprocessing to be to to be able to to extract objects from images so it's done in the played a major OK there's
illustration that you can do stereo electrification also with some satellite likely have
and also functions and what you need to do this uh
so so so the the the motivations is really due to to make multimedia interface between specialized tools to be able to combine them is a unified interface to be able to to construct and not by applying a remote sensing by applying all a part of your remote sensing by playing pool probably have also to combine with all the tools so uh and using those open source software will increase the number of algorithms and that of the available input and system doesn't OK so and and the cost what to is the better I teach the image processing library uh which is really but it is really based on and reusing the entire class to take of this library
OK compatibility will try to be optimal compatible as much but from a placebo uh is it that
pipeline of ITG that we we are using basically
internally we are able to make a number of these and able to process the images of any size I would say streaming the data I for each stream if the algorithm is ArfGAP ability for mergers threading there also an Internet of detector and general off detector which a lot to do this in a simple way so that's totally totally based on the United yeah I 2 aspects of and and that the biplane mechanism we try to stay as I said state of the art in the sense that we are using that the angle is genealogical algorithm which are used in medical image can and also is useful for remote sensing that also we tried to to to find the image-processing algorithm targeted on Geoscience that's and try to to to bridge the gap between the image processor was a receptionist and you end users also kinetochores loss of of remote sensing so with that by getting those algorithms and we try also to to work on what wasn't quite interested about the mean shift segmentation for algorithm because segmentation algorithm really helped to stream so if you want to want to to time your image processor segmentation and you can have some binary issues some kind things like this so well not only if want to work on the efficiency but also in some cases you will have to modify the algorithm and needs to be able to scale it with large data OK the welding
lots of statistics it's orexigenic stronger how to
use it to be so at the beginning it was at uh it's fully suppressed plus certain API so at the beginning to use it to be you have to be a surprise because the Republic was that the the the the initial way to accessible to be and use of layered imagery at that time well of course you can imagine that all the circuits because developers so we have the sum 2 uh upon the simplest + library to access the to accessible to be function so we've got but scope to the applications which provide OK uh more high-level functions to do segmentation classification etc. etc. and we but also dedicated your views radiation that the management of these arrays are common to the the that was developed for technical courses I think of that when
used fictional of the application graph 90 applications was saying now in quality that you can call from command line and graphical user interface space in the interface and without also by combining of all those applications OK so and
we've got multiple the so whatever it is not is not a GIS is really targeted that image processing what you can do with result about so we have developed this free fall to do something technical cost to teach remote sensing and the uh uh at the beginning of what you you can do a specific case to open the image is as large as you want is G Dallas of is able to open it you will be able to use related is related to the implementation of the there are some nice features and the a connection between open GL shatters individualize which is able to allow you to do some contrast enhancement in this kind of comparison between images uh and also it is not a GIS in the sense that we want to be able to to use images off electrified led like assumption to product that you gave you got from this have that you want to combine be able to these units at the same time on the same area I don't know high-resolution products live 1 which is not of quantified by default the default product that you can buy to Airbus digitally love is not up like 5 and want to be the internally to use sense on modeling into geometry to without doing the art of safety and actor recall to achieve staycation combined those images OK
and so what you get additions also available tutors and I was there during the clustering tool that to date the version of the the in the next section of this is the last version of what to be will be available so that is not an
exact binding of the applications applications without which are quite complicated and ought to be our simplified only those people's plates introduced but a key you can access to tools and applications what's new at all to be so I have made is not the last the last question was released in July so they have made a really is not simplify regions of of that as much as it was the 2015 we we we we need a measure of action Objection 5 because OK as I said there were lots of algorithms will to be uh a lot of applications but in term of modularity it was a little bit complicated to manage OK so we make it small modular following also the same principle that ITT was doing on that side so we have created uh suppressed this class C is being modules that people can understand what exactly what the module is done and groups etc. so we've got a better organization and that dependency management and a better way also to contribute to external motives to wanted the developers uh be but to be there was a fall followed by some joke about
that the the difficulty to install to be and all is those the the dependency of what it is is that it is not the case in my opinion we we're providing them with the scoreless scholarship w which is the 1 clique may configure project which is able to the don't instead of all the dependencies of ought to be and studied without any uh and packing of the configurations and you don't need to be an expensive although ought to be and all the dependencies to combine to combine OK uh have made a provisional as I said under the government's we have also ways I think from some some great progress on on spot we well I think working pretty pretty well on the fact that everything was open there was to be meaningless since the beginning of the project that we have now a Project Steering Committee official and all the governments all the decision made I think in the in the right way everything is public and it what it was on the other systems within the frame of the user viewing station and also because we want that their new organizations but to which which can be more current more involved in the development of what you did not only can s and and also it's who is working pretty well 5 . 2 we we have developed new new stop processings things so it was in the frame of the the availability of the something and 1 images there was since the beginning of license to to to to process something measuring what TV but less than the optical 1 we have we have done so some progress and added new applications and new functions for this week we've got at that time the support for fossil China what products medical but it would be a competitive you that of course and of course there we sentenced to and we've got also being able to get now official package there is none of shall package of what to be uh at indigenous yes I want to do really a special thanks to packages now their well not speaking to much of the of them held out there doing an incredible job so that simple and that for instance and began GIS is uh an incredible guy again well also making some developments in what about the 5 1 4 of when we want to say I will skip this discussion to to go directly to the last 1 but we have done in the last option in July 1st added this support to the geometry of something that was that was something which we which we using the data from something that 1 and of that of like fiber to get it from that that that you have and now we have to be we were liable to do this it is also possible in this function of 2 books of of these and so there is new new implementation of this off electrification algorithm will want to advertise that I would say that it's I think faster in inuktitut uh and that's why examples of new features of the that some slides the 1 is about what matters was discussing also the scanability you want me to be able to run the algorithm on the calls what is done in order to be used there is quite a new future of it I think it was done in the pretty clever way it's using the MPI so it is the message well I don't get it uh Misses which protocol interface so it's like a state of the art library to be able to exchange data in uh between jobs and between the uh between computer also that a key words which you can do now we want to be we will when you go you you yelled launching applications on STC computer it was around uh using multiple calls but a week it was 1 job it is really well using 1 1 computer now we we've got a new parameter you you you you you in style MPI you configure it in you it you either as the just addicted to launch the same applications at the MPI and appeared MPI parameter and we'll splits the scalar the the treatment on multiple computers and so it you you don't have to change the algorithms to to you to do some MapReduce all change the way you're processing the data of the same applications that internally we have added the possibility to 2 2 2 splits to speed the process
well working also on the classification cigar specification that the 1st part is to sample the data so we've got the data that we've got a lots of also of data to to rely on article isn't so there are lots of things to do an existing things which was difficult to do in order to be to to sampling take thousand polygons you want to randomly choose pick there's states all the exams you want to get some rules on the way that you are building up in your training set so we get new applications for this and
that it for the desire where would said to be the new features of the that trees uh my conclusion out of that 1 minute perfect uh mean menus of about difficult to say that is it's difficult to say
I think that's I see more and more interesting ought to be it so that the the number of of of of views are growing that the number of contribute to of use of application even contribute off we are participating in the decision process is I think increasing especially really since we we we have made the decision to to create these cases uh but we have
just more on the other hand because uh it that's
quite an issue in this part and the the applications for for for this kind of applications of its right now that he's going to have to increase I think a lot and we'll have I hope more and more use of OK some example of RO-TD was used with which is cool is that you've got to to be in ground segment you don't want to be which can be used to teach remote
sensing it can be be used also for the match analysis so that's cool it's also a problem because you have to to to manage all those different
applications and use of it it's it's cool to 2 to to be able to see that we've worked with a different type of users so we've got different projects we all starting to use it in the year following
them is a classic and can test again should map which is uh doing it which is done at the national level Support by specification which we're going to be done and possible with assumption intuitive along that take the product is in program the development is in progress and is fully based on what to and
that's and to stop thank you very
much so the question is you the militias in I have used by the body it is under 1 dual coding that you have recently into this in the visible so you have this idea using this are you are you using the these are the 2 we were using the these toolbox that it was more use of and in fact on this point we've got some people really we are we're were already using the TB with onset functional today that they want to combine this with the 1 that was that that some of bad experience I would say we is these are 2 works by by I think it is it's not a big issue that they wanted to within didn't want to get the new tools in there and the processing so they were really interested to put it to be and we're we're also we also trying to as I said doing some teaching of remote-sensing resulted in and not only optical data we really want to talk about that that we really what we really need to combine all those that sets and and we wanted we wanted also to be able to to teach 0 to process assumption that 1 there there are few functions in order to be to do this so we were also interested in that but to be able to to to have a full pipeline due to do this as the use of in OK the baseball field toolbox of was the name of the program it was combined cooperation at that time during the play and development between France and Italy so their coded of filled all the always was the at which they are combined and program so when you want to train users in the development of the satellite you want to get them to be able to use when the image will become becomes so it's the name was all fail and what I'll the is that the person would do the be 0 goal of the 0 that's why the graphical user interface is going on is that the flow of the information when you really believe in these images so what about the technology of combat use of the obviously know we are dealing with new role data but we are directly used for the processing on eyes don't we have directly exist in the real data for play and you've got 1 but off for the difference in the depending on the products spectral the we we will have 3 under and bonds and then we have well dealing with the real pixel values of course the back of the eyeball to export it of course so imagine these possible novelty informants of yes that's that's intuitively to can explore and it's using for reading and writing is completely using for the rest about invertible so don't sit well directly using those functions divided by means of the so the driver available on G that in all the format of schools I valuable noted it is a question and this is what the thank you for 1 another it was just the confirmation everything is open source all the coldest the transfer everything is open source on GPS we've got to go code review we are now integrating and again it was really that the resources the beginning there wasn't even a code review in the middle of the project in 2010 this uh during a presentation on Wednesday at tell that down discussion to moving from GPL to an Apache license that the that the classical feel that they want to do to to continue to do open source we resorted to uh even project like the the 1 on long so long classification of good picture of the and is based on a 2 D and it's also open source so different directly act as in cooperation the into 1 what is working in the authentication and certification house but they're expensive to do the altercation communities who's looking for all involved don't fall for the optical path OK we are we are using a PC and we is based on the post-implementation OK we are not aware that we are not developing the angle is that although the image quality you know whether that doesn't serve as dedicated for this for the geometry image quality what we are doing and what you need is to provide a tool which is able to read and identifying the right way and do the art of electrification OK for the sample of for instance we have re-implemented sort of to the occasion pipeline based on the use of these that would so but that is that is made the most and we've but also implement some income this also monitors in 2009 it's no we we are not using is that we could you this year that is supporting the efficacy for instance but there also when it was in some sense also lies about FIFO which is not working with the perceived which really is physical model okay which is much more complicated and is available with so we are using their policy some modelling which is completely different and there are also some sense also less now less number about that time there was a sense of which try using folds geometry about physical modeling it's much more complicated than some of them are available in with him and we are using it directly of what it is that there is doing the region the the computer doing the transformation to move from the images to the to the ground all the other way around so human assumes in decision so
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation
Computer animation


Formal Metadata

Title ORFEO ToolBox status report
Title of Series FOSS4G Bonn 2016
Part Number 165
Number of Parts 193
Author Grizonnet, Manuel
License CC Attribution 3.0 Germany:
You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor.
DOI 10.5446/20282
Publisher FOSS4G, Open Source Geospatial Foundation (OSGeo)
Release Date 2016
Language English

Content Metadata

Subject Area Information technology
Abstract Orfeo ToolBox (OTB) is an open-source C++ library for remote sensing images processing. It has been initiated and funded by CNES (French space agency) in the frame of a program named ORFEO to prepare, accompany and promote the use and the exploitation of the images derived from Pléiades satellites (PHR). Orfeo ToolBox aims at enabling large images state-of-the-art processing even on limited resources laptops, and is shipped with a set of extensible ready-to-use tools for classical remote sensing tasks, as well as a fully integrated, end-users oriented software called Monteverdi and also access via Quantum GIS processing module. This presentation highlights new features included in the last version of OTB, mainly major improvements of the classification framework which facilitate large scale analysis for automatic image mapping, new methods for SAR image processing as well as a recap of the main features added in recent releases. It further shows the current and future directions of the project and opportunities for interested developers and users. The presentation will be also illustrate by example of how ORFEO ToolBox is used at CNES in the development of operational products, scale over France, like annual Land Cover Map or snow cover product to exploit ESA Sentinel missions provided with a free, full and open data policy adopted for the EU Copernicus program. Manuel Grizonnet (CNES)
Keywords CNES

Related Material


AV-Portal 3.5.0 (cb7a58240982536f976b3fae0db2d7d34ae7e46b)


  557 ms - page object