Bestand wählen

Wartungsarbeiten am TIB AV-Portal

Am Dienstag, 24. April 2018 finden planmäßige Wartungsarbeiten am Backend-System des TIB AV-Portals statt. Während der Wartungsarbeiten wird unser Portal nicht zur Verfügung stehen. Wir bitten, die Einschränkungen in diesem Zeitraum zu entschuldigen. Sobald die Wartungsarbeiten abgeschlossen sind, steht Ihnen das TIB AV-Portal wieder in vollem Umfang und mit verbesserten Services zur Verfügung.

Vielen Dank für Ihr Verständnis!

verbergen
Merken

Open Source Work-flow for Surface Interpolation with Curvilinear Anisotropy

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
so my name's michelle Tobias I geospatial consulting UC Davis and where a new consulting group that works with and university to help other academics with a mapping into some until you'll that about a project that worked on about a year and a half ago and for open source workflow for surface interpolation of curvilinear and isotropy how the manifold to itself the anyway but with they were event is very so the
question really that this came out of was how can you accurately interpolate the depths for a meandering river channel is also applies to other things that kind of have that meandering tape and geography like coastlines that's my particular love but don't worry I work over channels today but you'll see of we'll see some
applications out of this cell this is the typical data that you get when you're working with River bathymetry and there's other forms of it but this one's kind of to the extreme so would they typically do is you have a boat in it has some sort of depth measuring device in this case this 1 had a sonar and array on it that took 4 beams as the boat moved and so they took this is actually this is the US GS publicly available dataset by the way all of the source of the end right and so they did was they put the on criminal boat in and they went up the river channel in theory at the deepest part of the river and so the issues so that it's hard to tell where is the deepest part on the surface of river that might be deeper murky and but was just pretend that that is the deepest by the river would not perhaps that and and then also typically what happens is not only you take that and so river channel measure then also people like to take cross sections so that's what you see on in sort of the zoomed in section there really dense section of data that was a cross section in this dataset we did a cross section every 5 miles per hour and so is not ideal and but this is typical of river bathymetry data sitting honesty in the the bigger chunk of data that you can kind of see where the cross-sections there really far apart and they're not super representative of the river channel itself and if I was going to collect data I my on spaces differently to get hold better data but working with what we've got an sometimes they do all kinds of crisscrosses and things like that but an that's typically what
we're going be working with an so if you've ever tried interpolation methods with this kind of data you know it doesn't work very well and what you see on the screen there is are the worst of the worst the idea of you so inverse distance weighted typical decades have IW doesn't always doesn't usually work that well for most things anyway but with river data is particularly bad so if you pretend that the dark areas a deep in a lighter areas a higher you can tell this river will not flow her and all but there's like some kind a massive sinkhole group and the northern section and so why doesn't it work on part of the problem is that the data density is really variable in this dataset you got it up the middle of the channel and you've got these 5 mile apart cross-sections I'm also the is really highly an isotropic so what that means is that the points that are next to each other and if it was isotropic they would be closely related so point x each other in any direction should be similar but I'm if it's an isotropic that means that and data is related to each other only in 1 direction so and for example like yeah person tetradymite so I can move around but so did in this direction would expect to be similar to each other on the problem is that that doesn't apply when we moved a different section of the river like right here we don't expect the anisotropy to be in the same direction everywhere in the river so we have variable and isotropy direction supply regular gene with and isotropy because the direction of the isotropy changes so you ever tried to apply regulatory into this kind of data set and it works in some sections but not in others especially it was a really curvy dataset I should mention this is the emmit river actually the the dataset so it's this is the part that runs the Portland but the just happened be the
dataset found or so how do you make this work this is a picture of my research note that it's color coded UA and so what I again for this particular data set all give you good big scary overview and will break it down and so the data came in and 4 so in text files how to write a Python code to get the text files into 1 dataset that I could put into Q GIS which is the orange color and after that we did digitizing whole show you how that works in a 2nd and then brought the data posters and then us smart you just more and posters and they're brought in are to do the creating and then back to posters and then back to GIS so that's why this is a presentation of a workflow because we're dealing with all kinds of open source tools smushed together and it is something that actually works which is cool
and eventually will be nice automate this in some way so that the user could just put in some input and then magical happen but also you have make magic happen here and so the overview is you need some kind input data once you have the input data and then it will work actually do a strain out the river where make all the anisotropy going 1 direction so we can do the grading on it we don't do fancy creating and just regulatory and then an were going to do that interpolation those used probably treating batteries for this and then we're gonna read banners estimates back into that for river shape that were more familiar with
at which sounds really scary history artist reserve organisms I was like 0 no the states of the go mouth and it is and so in Q giants need the input data for this particular set of data and you need depth measurements which you're going to get from the field and work or from the USG US online and you also need a river polygonal lines said the blue is in the middle and then you'll need a reference lines that offset from the river and the off cell line is really important for some of the math we need to do later so then you can take that river polygon in a you wanna make a regular grid over the whole river and you can use that river grade 2 clips of the river polygons clip the grid just because you don't need the points outside of the river and and for processing it it slows the weight of the of points you not in use the ANSI muscles cloak amount to start at or so therefore well we 1st has and so the
next part once you have all the input data were in a string of the river but this is based on an a paper by more widely in 2005 and they actually there's a team of people that wrote a plug-in that does justice for our GIS interviews that except it was an old version I have to get my IT people to install the version of Ottoman figure I'm explain work so this actually this process was easier to leave are not an so it is not based on his paper so now you can use a river reference line and you're gonna reference every single 1 of your points in both the field sample data so the things that came out of the and then also you River grid the regular grid that you made those all need to be referenced to the river reference line so we're taking things out of x y coordinates latitude longitude were transforming them into reverse space so we need to calculate how these points lined up with the distance all along the river line so that Kirby distance and if you're walking along the river and then the distance from the river so we're taking a perpendicular measurement from the river to a point and that's going to be the end where the distance along the river is s the the
so how would you that is and it's pretty straightforward in posters and I see that now have their work happened and but so they're really key part here is that once to get your and coordination and uses st distance command in and posters and you give it your line geometry and then you also get the point geometry so whatever set of points you to reference and to get that distance away from the reference line and then to get the escort which is the distance along the river so that curvy measurement and that st line locate point himself and you put that in you give that also the line geometry and and do some fancy now wanted to get the distance along it and and then that teacher S on so this is the table for an the point so these are the and the bathymetry points you throw those a new reference those to the S and n coordinates the and then you also do that with your regular grid points so similar mass similar and query in post just to get that out see new those on you'll see
later on in the next slide you'll see how all lined up and becomes nice and straight and manageable so once you have those straighten points it will look like 1 big call on and so the big the mass I you need a fit a very ground this is when I brought it into ah I with and I think for the client they did this for actually and abusing arc an or was the thing that was going go quickest obviously running out of time for presentation at and we could do it you can do this at any GIS they're comfortable with C take your starting points in S and N accordance you bring it into a GIS and that can do creating a need to fit a very your data in order to do that creating so in this case this is where you stand on a the G. stat package in the very grammar it probably to be fit better but again time was of the essence of this actually works out pretty well and so you can you pretty much have to play with the data and figure out what the best model that is going to be for the very ground and before you can use creaking so then you're going to put
that his nights in a your input that and variogram model into you're creaking formula so that you can figure out what your estimate for your points it is an this particular ad tool in estimated doesn't make a roster even all looks like a roster it's actually points and it's using that regular point grid and it makes the estimates for each of those points in the grid rather than making a roster which you would then have to sample the points but I in general was in the end you 1 end up with some points so you can see it the strain out it converge projectors not displaying it wonderfully but on when it's strain I don't quite see which you might expect in a Weber in terms of depth but promises gonna work out the printer much playing with it I realize that kind of make it look like a river in this particular a step was not going to happen so that promise it works out now
so now you have these straightened estimates
nice straight River column it's not
time to read bend it back and the now the cool thing is if you keep your but ideas for each of the points in the grid you don't have to do the math to read transform I you can just do a join see join your point estimates the straight point estimates back to the curvy once and then it just fits very nicely however you could do the math backwards and and put it back where it goes but skewed the ideas and it's a lot easier to found this out from experience and this there's no need to do MAP backwards and so this is a simple join query and PostGIS and so just enjoy and based on the ideas and to get all that back right so
in the end you end up with I know looks like a roster on the and on the left of the screen but that actually . 38 so many of them there is having this were 5 meters apart and the so many of them they look like a restoration serotonin q and then I turn it into a roster and so now you can see it actually looks like you would expect a revert to look rats that IDW I showed you the big blotches of color looks cool but not really river and so you can see the debts here at the deeper spotted darker in this picture you can see how the river curves within its channel I you can see that it is deeper on outside of the curve which is where you would expect right that's the the cutting side of the river bank and then it shallow warm inside which is where the depositing is happening of the sediment so that this estimate is actually working all lot better already then then the IDW or even any kind of creaking would and the so this is really exciting even living maybe a very and that's not fitting quite as well as we would like you still get pretty decent and results with this so that's really nice or I'm not sure you can see the contours at all on the projector and back to the side of the slide actually has the contours and it's even looking out at all on an so I'm really pleased with how this came out if you zoom in a certain sections of the river there are places where you wanna go in and kind of work with the media and the filter they did a little bit better even with the sun I sometimes get points that are anomalies so you want a filter that probably the and also the ends of this really don't work and you don't have enough points and so can drops off everything transferred 0 at the end of the rivers to guarantee plan for clipping man out the the
finish literally I think that's OK the and
so I guess at this point I'm willing to take any questions you might have and I'm probably skipped over something important and so it happened click so feel free to ask so I think we have a microphone to the the what Michelle that's a owl was warmer if you do this to me over the future than the other algorithms like our radial basis functions were or anything else to do the interpolation I have not for the project for this and for some reason we work with clients they get hung up on the train their variances and they need is some form of creating by adding really could you really could work with different and interpolation methods on this and just drop in and you know different things or use something that was gonna work better from the beginning and all of this bending stuff here the question for this task you book so you audio auditable process so you can just throw some differences this the river no makeup or drugs or under traditional Brussels I have not automated and that would be really awesome to do so if anyone wants to help me with that would be excellent and I don't currently have my jobs based on what funding to work I I have funding for that right now but if anyone wants to work on it I'm more than happy to to can help with that it would be really nice it if you look online and like stackexchange and things like that there's a lot of people asking questions about how to do this in the answer it until now has been open can go back a few steps in Ottoman downloaders plugin and then you know and that's not a great solution for a lot of people the next of this was open source the and when you're talking about the the yeah the reference length the room a and that's slide it would like the reference line to the side of the river
was the reason for not so uh
looking reference land on the center the VA's
something yeah you could I I think I think I set it off to the side that made the mouth or will be easier for me to to work out my head but you could you could put it down the center and then I imagine you've probably end up with and some points that had a negative reference now fine and arms and 1 on your phone bill the photo over the years it probably would have wanted and done this before and in the sharper curves you end up with points that I overlap each other in ways that they wouldn't normally so yeah that is a problem that's why you definitely you go back and see what the end result is minus places that are kind of messy and in the end of end up in calling data in certain areas to make it work better yes rely post-processing on this this so nice and pretty this workflow but it definitely takes a lot of work to make it work right now yeah the show up in the the drying where you straightened out the river it looked like it was
even uh and even grid all the way
down but the the river itself I mean is different sizes so did you just add extra columns and then masked those back out and these there an I think it just looks like a straight here because of the way that part of the scale I think the zoomed in on sections you would see there's need missing in certain places and but it's so long that with alright it's you can really see in an input that bounding box around it that's not the river channel itself so then you just masked the part that when you wrong no just use the rate of whatever wherever the points from the grid land you just use those so we're not adding adding anything else in that's broken in question Ch and but reshape come a lot of big words and are further so if we think the
Mapping <Computergraphik>
Krummlinige Koordinaten
Interpolation
Open Source
Gruppenkeim
EDV-Beratung
Projektive Ebene
Grundraum
Topologische Mannigfaltigkeit
Ereignishorizont
Computeranimation
Open Source
Selbstrepräsentation
Magnetbandlaufwerk
Zellularer Automat
Raum-Zeit
Quick-Sort
Physikalische Theorie
Computeranimation
Streuquerschnitt
Dichte <Physik>
Bildschirmmaske
Rechter Winkel
Mereologie
Garbentheorie
Einflussgröße
Punkt
Gewicht <Mathematik>
Open Source
Mathematisierung
Inverse
Gruppenkeim
Elektronische Publikation
Kombinatorische Gruppentheorie
Code
Computeranimation
Richtung
Streuquerschnitt
Dichte <Physik>
Menge
Interpolation
Regulärer Graph
Flächeninhalt
Rechter Winkel
Mereologie
Garbentheorie
Kantenfärbung
Abstand
Touchscreen
Shape <Informatik>
Punkt
Gewicht <Mathematik>
Selbst organisierendes System
Mathematisierung
Zellularer Automat
Prozessautomation
Ein-Ausgabe
Polygon
Computeranimation
Gradient
Richtung
Datenfeld
Interpolation
Menge
Einflussgröße
Gerade
Lesen <Datenverarbeitung>
Aggregatzustand
Punkt
Prozess <Physik>
Versionsverwaltung
Abfrage
Ruhmasse
Ähnlichkeitsgeometrie
Ein-Ausgabe
Räumliche Anordnung
Raum-Zeit
Computeranimation
Datenfeld
Menge
Regulärer Graph
Mereologie
Stichprobenumfang
Abstand
Figurierte Zahl
Koordinaten
Einflussgröße
Gerade
Tabelle <Informatik>
Zeichenkette
Schätzwert
Punkt
Formale Grammatik
Ruhmasse
Statistische Analyse
Ein-Ausgabe
Kombinatorische Gruppentheorie
Term
Computeranimation
Rechenschieber
Codec
Informationsmodellierung
Client
Regulärer Graph
Stichprobenumfang
Ordnung <Mathematik>
Beamer
Figurierte Zahl
Fitnessfunktion
Mapping <Computergraphik>
Schätzwert
Punkt
Verbandstheorie
Mathematisierung
Abfrage
Gerade
Computeranimation
Resultante
Schätzwert
Bit
Punkt
Automatische Handlungsplanung
Snake <Bildverarbeitung>
Computeranimation
Programmfehler
Rechenschieber
Hypermedia
Meter
Garbentheorie
Kantenfärbung
Kurvenanpassung
Beamer
Metropolitan area network
Dicke
Subtraktion
Wellenpaket
Punkt
Prozess <Physik>
Open Source
Radiale Basisfunktion
Computeranimation
Rechenschieber
Task
Client
Bildschirmmaske
Algorithmus
Interpolation
Prozess <Informatik>
Rechter Winkel
Projektive Ebene
Varianz
Gerade
Resultante
Punkt
Flächeninhalt
Digitale Photographie
Kurvenanpassung
Computeranimation
Schreib-Lese-Kopf
Zentrische Streckung
Punkt
Quader
Mereologie
Wort <Informatik>
Garbentheorie
Ein-Ausgabe
Bitrate
Computeranimation

Metadaten

Formale Metadaten

Titel Open Source Work-flow for Surface Interpolation with Curvilinear Anisotropy
Serientitel FOSS4G 2014 Portland
Autor Tobias, Michele
Lizenz CC-Namensnennung 3.0 Deutschland:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen.
DOI 10.5446/31682
Herausgeber FOSS4G, Open Source Geospatial Foundation (OSGeo)
Erscheinungsjahr 2014
Sprache Englisch
Produzent Foss4G
Open Source Geospatial Foundation (OSGeo)
Produktionsjahr 2014
Produktionsort Portland, Oregon, United States of America

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract Traditional approaches to interpolating a surface from point samples often assume that if anisotropy is present in the data, it is along a straight line. However, this assumption breaks down for many data-sets, such as river bathymetry, elevations near coastlines, or levels of pollutants along roads. I will present a work-flow using open source tools to make use of the curvilinear anisotropy present in many data-sets by transforming the curved data-set into a straight line (QGIS and PostGIS), applying an interpolation to the transformed data (SAGA, GRASS, or other program with interpolation tools), and finally re-bending the interpolated surface back to the data's original shape (PostGIS).
Schlagwörter interpolation
QGIS
PostGIS
GRASS
SAGA

Ähnliche Filme

Loading...
Feedback