Activity Map from space: supporting mine clearance with Python


Formal Metadata

Activity Map from space: supporting mine clearance with Python
Title of Series
Part Number
Number of Parts
Cammarota, Giuseppe
CC Attribution - NonCommercial - ShareAlike 3.0 Unported:
You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal and non-commercial purpose as long as the work is attributed to the author in the manner specified by the author or licensor and the work or content is shared also in adapted form only under the conditions of this license.
Release Date
Production Place
Bilbao, Euskadi, Spain

Content Metadata

Subject Area
Giuseppe Cammarota - Activity Map from space: supporting mine clearance with Python Removing UneXploded Ordnance (UXO) from minefields at the end of a conflict is a very time-consuming and expensive operation. Advanced satellite image processing can detect changes and activities on the ground and represent them on a map that can be used by operators to classify more dangerous zones and safer areas, potentially reducing the time spent on field surveys. We exploit space-borne radar Earth images together with thematic data for mapping activities on the ground using numpy, scipy and gdal. The Activity Map generation process to be shown will be implemented using IPython Notebook.
EuroPython Conference
EP 2015
EuroPython 2015
Satellite Ocean current Slide rule Mapping Spacetime Process (computing) Physical law Complete metric space Open set Data mining Goodness of fit Computer animation Information systems Vulnerability (computing)
Area Slide rule Sign (mathematics) Process (computing) Dataflow Source code Bounded variation
Point (geometry) Satellite Mapping Dynamical system Weight Medical imaging Mathematics Mathematics Single-precision floating-point format Videoconferencing Cuboid Metropolitan area network Area Metropolitan area network Simulation Spacetime Texture mapping Information Projective plane Planning 19 (number) Arithmetic mean Process (computing) Computer animation Network topology Object (grammar)
Metre Mapping State of matter Scientific modelling Direction (geometry) Multiplication sign Cloud computing Insertion loss Mathematical analysis Event horizon Theory Storage area network Number Image resolution Medical imaging Mathematics Hypermedia Average Energy level Gamma function Subtraction Units of measurement Position operator Area Metropolitan area network Time zone Process (computing) Scaling (geometry) Information Computer Least squares Grass (card game) Maxima and minima Word Network topology Personal digital assistant Mathematics output Energy level Figurate number Object (grammar)
Histogram Pixel Mapping Thresholding (image processing) Field (computer science) Medical imaging Mathematics Root Core dump Square number Data structure Gamma function output Extension (kinesiology) Atomic nucleus Social class Metropolitan area network Projective plane Sound effect Functional (mathematics) Inclusion map Arithmetic mean Computer animation Function (mathematics) Mathematics output Library (computing)
Building Mapping View (database) Covering space Combinational logic Water vapor Function (mathematics) Mereology Rule of inference Medical imaging Mathematics Hierarchy output Subtraction Social class Covering space Area Metropolitan area network Time zone Texture mapping Process (computing) Information Limit (category theory) Inclusion map Data mining Arithmetic mean Computer animation Mathematics Function (mathematics) Optics output Analytic continuation Object (grammar) Game theory Resultant
Point (geometry) Mapping Multiplication sign View (database) Geometry Source code Design by contract Mereology Discrete element method Medical imaging Energy level Software framework output Position operator Area Turing test Product (category theory) Texture mapping Spacetime Process (computing) Information Projective plane Fitness function Metric tensor Data mining Message passing Process (computing) Computer animation Software Function (mathematics) Eigenvalues and eigenvectors Order (biology) Configuration space Energy level Remote procedure call Library (computing)
Medical imaging Population density Process (computing) Computer animation Angle Geometry Mathematical analysis Function (mathematics) Object (grammar) Orbit
each what the good afternoon everybody of current work of Brooklyn solution is a company that works with the satellite imagery and general with geospatial data and is located in so and what about you do not supporting many years with Python so there is for the topics we are going to to talk about the theme of what the material the among these whitey it is from space what's mind here in the mind and what what so this stuff from the end was by does anyone knows what might what you you but as the yes because I put this slide just for the sake of of completeness so let's go to the monkey and is the process of searching and removing landmines from the ground effects some weakness trouble weakness the original 1st weakness mine mine clearance is law
is a very slow process you're this slide this huge area you have to clean from minds and nobody will tell you what the minus sign on national army and not so collaborative installed in on nobody the likes that the deployed landmines so yeah is usually you have to search for my source then these Eugeniusz you have to check for landmines step by step they export rise fall very slowly some very slow process you are always in danger slope when you are in school by what you think it's in my top really slowly try to clean the the area around finally variation and trying
to remove but there is only for obvious reasons another problem mine clearance is as
I said is that in general injuries there more often than you might think even now we've been professionals this is why at some point there goes about using many months to search for mind to sneak them and at 1st they thought using boxes but sometimes you just the the weight of 1 Q I now announced to the my so is then generalized also so a couple of years ago they started using the for searching in mind the way just uh a few other announced and they have a brilliant plan so all these things but and there are more Q for years searching for cheese and talk about a topic of that you that during this project actually so there is the era of thinking that you this this job but we can do with if we go in this space that we there is
plenty of satellite would use capability huge probability is this is a simulation of the dynamical solution possible tyrant it produces rather images that have the 1 about the news that for this job the constellation is provided for something like and the there are a lot of between 6 and 700 kilometres from the ground so it's really they would be safe from landmines and they complete a single object in nineteen mean you might mean so every 90 minutes they be the novel that on so the other fast well this seems to cover to fix the problem of landmines so as I said they produce another images that are and this is a coherent sentence or so you have all the information about the competence of missing signal and the face of this thing that we are going to use both of them the that's going to what an activity map so I think the man these are mathematics information about all this about the activities that goes over a certain area for activities is men like everything that the produces changes on the underground they mean if you read the new alongside so if it is of what trees are the best practice if you use a role the 4 year olds I used you can say if there is a video you can think you can see from something like so classical of the
tumor classified status based on the level of activity In this case the green area I wonder at the and that's why I have the units so we will but that lead to loss of the word the area where there is no activity this is very important for landmines because 2 if there is a DVD on the ground you can say that the basic model as the minds of the we living last so if this is a very good the direction and the lot of the process is is an example of an area you know Bosnia I believe of 5 you don't need the by kilometres away about the twenties with you me that was to they would take ages which completely I would say it among and but we can reduce this time just overlaid on top of these and I think of so you can but immediately what you don't need to go to search for landmine where it's going because something as something here so you know that state and you can focus your efforts on searching for minds on the other zone where there are where there is no information about it what do
we use to produce an amount another images associated with this this is an example of an object of a number that images this is banned by thank you don't need what with the resolution of 1 meter so this is what was talking and 8th of July 2 thousand 415 it is worth uh dollar figure is a city in a in the a positives evolving then we have least a 2nd image just to compare with the 1st is 1 on the 3rd was quoted 15 In this way you can search for changes then using features of these 2 images you can compute what we call the interferometric coherence that basically there when things have changed between the 2 images the the darker the peaks in these the most the most of the data as changed this is the scale so where these wide so there is a city there so they're mostly being things that we keep in the same between the 2 images why what is that is mostly rigidity the various so leaves uh most and you can see that things have changed so it's pretty and things that used to compute that theory cannot is what we call coherent changes so this so that these are the changes that that you can see just comparing the object so the interferometric coherence sexually ruined like before it was common computing it was than using the faces of the you in this this is an average of we call complicit you might use it means that in the event that may it has been put on the 1st image is the 1 token on the 18th of July and in the green and the blue layer as the input the 2nd images so if there are 2 opportunities for without being the same you would have seen totally black and white images but since you see some reddish bark and some science corner that means that in the zone where is that for example in the amplitude of the images in the media in different layers uh these ideas right there in the sciences that's why you see that so in this way you can see some changes already comparing the amplitude then if we did the same with another kind of complicit images so it's just using the the coherence in the Brunei instead of over the you can produce this coherent changes that does this in more detail that does not think that information produced these to produce what we call a land use and highly of fundamentalism and but that's you in a semi-automatic way way what these on the on the data you are seeing for example you can that the underwater areas where the you run area are where the these Green grass tree and we use a lot of this then to produce an activity sure the state that's the producing
I 1st of all once you have the data you have to stretching benefit that so just that just using the 5 years and some searching function like as well as square root function normally found a linear fashion and then and fish that with function from by the library here what we do is basically be the histogram of the images then decide where to cut the dynamical this is by deciding what left bound that has to be aware that I want to be and then compute the stretching in this dynamic just to show the effect of this that this is there a debate that input as X stop before or after an extension of the view that
this is the 2nd major is what is being structure of the so every feature in the images is morally organized what you can see fields roads our and 3 then the 2nd step that the the changes classification this is a really simple called that would say we just 1 instruction in the formal within classifier on the other hand the images we just decide where the threshold for deciding which class every pixel belong and then we launched lunch this formal so a as an example of these this is the input images of this step is the quality and the uh the from the inquiry and socially recall it goes from 0 to 1 meaning 0 nucleus of all set so that is totally totally has changed the 1 used for the core you everything just give the same between the 2 images after because they're classification projects you just have a classification because the images here 6 classes where there is no doubt that outside the images then it is
the 1st that that is the combination between the much because the changes classification and the land cover land use that something about it before it is because if you can be that the change in this as I went over a rose is a different meaning these because of an underwater or if it because up all over the object in the area so this is just a very simple I was a new classification that take these aspects in the top so there are for example classes like 3 that is what they're not always the same the bonds with landmines and water or only in the view among classes the means that of buildings on 1 with the land mines and the so and then again you you just have this new classification so that they that takes into account the systematic different mapping information of the and this the game is an example of 1 of these steps data so this is the 1st input that their careers and the changing images together with the change of
land use let me and then this is the output is there a lot of this is the area of that activity among us with a Pacific region so you can see that the sigh and rules for example having an active so you have to search then while the bed that zone hierarchy so we need to go there for search for landmine and then there is this yellowish part of that is where the process where applying is not topic of 1 that is the result of some intrinsic limits of the province of and it can be solved using some other kind of sense like optical or infrared money infrared and so on that can beyond that can be added to this problem I'm not sure you can now what you can do it
then the final step once you have the metrics with the amount just as to what the geographical information and library brilliantly his job very simple called so this is the step to cross from the just my fits with the value and what associated to produce so a layer which graphical information that can be used by professionals the use of grammatical or every itself the software these in commerce but reduced produce just of course the other concluding my the we just find just find some final remarks yeah amount of mine maps so a 2nd level problem that is supposed to have at to up all my years professional trying to reduce time and some money invest in this process and to reduce the danger of doing such a job is that the the map is a product that was developed in the framework the European Space on European Space that project called the space assets foreigners in their mind and this atom project is going to be tested on any of the next September inside you the kind what B which of questions at the end of the process so my
question is less about Python and more about the data sources how widely available are these remote data source did you have to I guess make special requests to get coverage of that area the data from every pass every day some were available what do you mean that there is something like images of while these that since you get a contract with the company that produces these data you just lost them you place an order and fixed so that's something I would pass over the area of there is just in the data and then we will passing every 24 hours In the 2nd part of the because just because suddenly orbiting around the entities that we have responded to this that is so of that out there would be the same configuration again you need to say something good to compute the coherence of the images because you need to face so you need that the something that has be in the same position to get the same geometry of the images you could like uh meaning at least for might use on the on the same area but they will not be on the same geometry we what with to that light is looking at the from the point of view then
often I think that that might
mean the orbit is called and of thinking simplicity should not be a slightly rotated so you would just see these things being the same idea but just with another angle so if you want just Lucene the NLU think we did at the success the 4th densities it depends on the object and that none of you go from the output of the more process US but you just can't do the phrases analysis using the images that have been thinking with the same geometry so that's why you just wait for 24 hours for the 2nd and the images the what questions do basic tank features of the


  545 ms - page object


AV-Portal 3.9.1 (0da88e96ae8dbbf323d1005dc12c7aa41dfc5a31)