An Automated, Open Source Pipeline for Mass Production of 2 m/px DEMs from Commercial Stereo Imagery
This is a modal window.
Das Video konnte nicht geladen werden, da entweder ein Server- oder Netzwerkfehler auftrat oder das Format nicht unterstützt wird.
Formale Metadaten
| Titel |
| |
| Serientitel | ||
| Anzahl der Teile | 188 | |
| Autor | ||
| 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. | |
| Identifikatoren | 10.5446/31715 (DOI) | |
| Herausgeber | ||
| Erscheinungsjahr | ||
| Sprache | ||
| Produzent | ||
| Produktionsjahr | 2014 | |
| Produktionsort | Portland, Oregon, United States of America |
Inhaltliche Metadaten
| Fachgebiet | ||
| Genre | ||
| Abstract |
| |
| Schlagwörter |
00:00
MeterMathematikPunktprozessFlächeninhaltQuaderEinfügungsdämpfungDatensatzDialektAuflösung <Mathematik>FlächentheorieURLMotion CapturingNumerische MathematikEinflussgrößeAnalytische Fortsetzungt-TestSatellitensystemPixelMengenlehreEinfache GenauigkeitProjektive EbeneMinkowski-MetrikTaskOrbit <Mathematik>Formale SpracheRechenschieberCASE <Informatik>Registrierung <Bildverarbeitung>Computeranimation
02:54
MereologieDialektPolygonVakuumpolarisationStereographische ProjektionTermPolstellePackprogrammKontextbezogenes SystemSatellitensystemGreen-FunktionNichtlinearer Operator
04:18
RFIDMultiplikationsoperatorHypergraphCluster <Rechnernetz>KoeffizientEndliche ModelltheorieWorkstation <Musikinstrument>VerzeichnisdienstProgrammierungWeb SiteBitOpen SourceFunktion <Mathematik>SpeicherabzugSystemprogrammGerichteter GraphGeradeBinärcodeGanze FunktionMengenlehreProgrammbibliothekDivergente ReiheBildverstehenGüte der AnpassungSpezifisches VolumenTermSoft ComputingParametersystemBinder <Informatik>Metrisches SystemPunktwolkeModulare ProgrammierungHalbleiterspeicherWrapper <Programmierung>Rationale ZahlPolynomQuick-SortPolarkoordinatenLeistung <Physik>PunktComputeranimation
07:10
Endliche ModelltheorieDigitale RevolutionAutorisierungPunktwolkeFehlermeldungFunktion <Mathematik>Divergente ReiheTextur-MappingTesselationFilterung <Stochastik>Produkt <Mathematik>TermARM <Computerarchitektur>SystemprogrammPunktDiskrete-Elemente-MethodeComputeranimation
07:55
Funktion <Mathematik>SystemprogrammSpannweite <Stochastik>Workstation <Musikinstrument>Produkt <Mathematik>MeterTermVerschiebungsoperatorURLTextur-MappingFehlermeldungKreisbewegungFlächeninhaltSchmelze <Betrieb>Güte der AnpassungNichtlinearer OperatorZehnMengenlehrePlotterPunktGeradeTranslation <Mathematik>PunktwolkeRegistrierung <Bildverarbeitung>Numerische MathematikMultiplikationsoperatorStandardabweichungDialektKeller <Informatik>Rechter WinkelSplineStarrer KörperBruchrechnungStichprobenumfangAdditionDifferenteTopologieDemoszene <Programmierung>Überlagerung <Mathematik>GamecontrollerPixelGauß-ProzessKnotenmengeDiskrete-Elemente-MethodeProzessautomationMultiplikationPolstelleMedianwert
11:23
GeradeRandverteilungMultiplikationsoperatorZeitreihenanalyseDichte <Physik>MeterKugelkappe
12:14
MultiplikationsoperatorZeitreihenanalyseDichte <Physik>FrequenzDiskrete-Elemente-MethodeComputeranimation
12:39
FlächentheorieTwitter <Softwareplattform>MeterLinearisierungSummengleichungDreiecksfreier GraphEinfügungsdämpfungPixelPunktMultiplikationsoperatorKlasse <Mathematik>URLFormation <Mathematik>FrequenzAnalysis
13:47
Produkt <Mathematik>EchtzeitsystemMultiplikationsoperatorComputeranimation
14:11
MeterPixelMereologiePhysikalischer EffektEinfügungsdämpfungFokalpunktComputeranimation
15:05
MeterPixelTextur-MappingDiskrete-Elemente-MethodeFlächeninhaltMathematikProdukt <Mathematik>Komplex <Algebra>Logischer SchlussKlasse <Mathematik>Rechter WinkelDichte <Physik>Thermodynamisches GleichgewichtWürfelMinimumComputeranimation
16:17
MultiplikationsoperatorSchmelze <Betrieb>MathematikZentrische StreckungWeg <Topologie>Produkt <Mathematik>Textur-MappingMinkowski-MetrikURLEinfügungsdämpfungComputeranimation
17:01
PixelDiskrete-Elemente-MethodeProdukt <Mathematik>MultiplikationsoperatorGeschwindigkeitDerivation <Algebra>VektorraumFlächentheorieEinfache GenauigkeitAuflösung <Mathematik>URLWeg <Topologie>Korrelationsfunktion
17:39
AnalysisKonzentrizitätSichtenkonzeptQuick-SortURLMultiplikationsoperatorOffice-PaketSkalarproduktPolygonAggregatzustandMehrrechnersystem
18:37
Güte der AnpassungProdukt <Mathematik>SichtenkonzeptTermMultiplikationsoperatorAuflösung <Mathematik>VersionsverwaltungDatenparallelitätKorrelationsfunktionTopologieZoomRechenschieberComputeranimation
19:32
ZoomRechter WinkelMehrrechnersystemDiskrete-Elemente-MethodeDifferenteMeterTermBildschirmmaskeTafelbild
20:35
Patch <Software>EinflussgrößeWald <Graphentheorie>TopologieGraphfärbungTextur-MappingBitWort <Informatik>MathematikMultiplikationsoperatorRauschenMeter
21:33
MeterMathematikEinflussgrößeSpezifisches Volumen
22:05
FastringArithmetische FolgeResultanteMathematikZustandsmaschineInformationsqualitätArithmetisches MittelComputeranimation
22:42
Produkt <Mathematik>Endliche ModelltheorieMeterSplineGauß-ProzessPixelPunktDichte <Physik>MultiplikationsoperatorAbgeschlossene MengePunktprozessSondierungDiskrete-Elemente-MethodeMosaicing <Bildverarbeitung>Auflösung <Mathematik>Dienst <Informatik>ProgrammierungProzessautomationTemporale LogikAggregatzustandZeitreihenanalyseGamecontroller
24:08
InformationSystemprogrammAuflösung <Mathematik>MultiplikationsoperatorTermEinsFunktion <Mathematik>Lokales MinimumMereologieProdukt <Mathematik>MeterKartesische KoordinatenDiskrete-Elemente-MethodeFlächentheorieProzess <Informatik>PackprogrammLuenberger-BeobachterFlächeninhaltAggregatzustandBenutzeroberflächeMultigraphMinimalgradElektronische PublikationValiditätStapeldateiMathematikAutomatische HandlungsplanungSchießverfahrenWeg <Topologie>PunktprozessMengenlehreKnotenmengeHypergraphVorlesung/Konferenz
Transkript: Englisch(automatisch erzeugt)
00:02
All right, I think we should probably get started here. It's 3 o'clock. So I have a lot to talk about. I'm very excited about a lot of the things I'm going to show here. My name is David Sheen. I'm at the University of Washington, and I'm working with a number of people who have been instrumental for this project, including Claire, who is here in the front row.
00:22
So this talk, I'm going to give some background. I'm going to talk about the massive amount of data that we have, talk about our processing tools, some co-registration and accuracy analyses that I've done, and then show a lot of pretty pictures, which I think you guys will enjoy, for places all over the planet.
00:41
So this is my one-slide summary of what we're doing here. Essentially, we're taking stereo satellite imagery, and primarily we're using Digital Globe, WorldView 1, and WorldView 2 data. These satellites are incredibly nimble and precise, and in a single orbit, they can acquire in a long-track stereo pair. So they can pitch forward and acquire an image, and then about a minute later, pitch
01:02
back and acquire an image of the same spot. These images are half-meter per pixel. When we receive them, it's a 17-kilometer swath. So it's 17 kilometers by however many kilometers you want in the other dimension. And we've developed a pipeline to take those images and automatically produce high-resolution continuous digital elevation models.
01:21
And then we're doing this over and over and over again for some high-priority areas, like this location in Greenland. So what does half-meter-per-pixel imagery actually look like? This is an image acquired in Greenland in 2011. This is the Greenland ice sheet here. Those black things in the inset there, those are surface meltwater lakes.
01:40
And that little red box is this larger image here. And you can see there's a stream channel here. Some of these darker areas are – there's meltwater on the surface. And then there's these strange bumps, and those are actually mountain tents and these 8-by-8-foot tents from the 2011 Woods Hole and UW camp. So that's the kind of resolution we're working with here.
02:03
I'm a Ph.D. student. I'm in the Earth and Space Sciences Department, so I'm tasked with doing the science. So the motivation for a lot of the work that I'm doing here is that these repeat measurements of surface topography can capture ice loss and, in some cases, ice gain. But really, they characterize the processes that are driving a lot of the changes that
02:20
we're seeing today on the ice sheets and elsewhere. And we care about this because all that ice that's leaving the ice sheets, it ends up in the ocean. And every year, sea level is rising 3 millimeters a year, and about 1 millimeter of that is from the ice sheets. We want to know where that's going to go in the future. We know it's going to increase, but we don't know how much. We can also learn things about regional climate change by studying mountain glaciers.
02:42
And something that the ASO talk earlier really hammered home, we can – if we know elevation of seasonal snowpack, we can learn a lot about water resources, which is really critical for the Western United States. So in terms of coverage, here's what is available as of June 1st in the Digital Globe Archive.
03:01
Every polygon here is a long-track stereo pair. So there's 175,000 of these, roughly. And when you look at this, you can see there's excellent global coverage. But really, the polar regions are extremely well covered. And part of that – the reason for that is that there is now six – now that World
03:22
Every 90 minutes, one of these is flying over one of the poles. So there's a lot of opportunities, but there's also essentially zero competition. There are very few customers that want to take a picture in the middle of Antarctica and pay Digital Globe to do that. So fortunately, the federal government has an arrangement between the National Science Foundation and the National Geospatial Intelligence Agency and a few other agencies to make these
03:43
data available to researchers. So these are the – how many people have actually used – have used a polar stereographic projection before? I'm just curious. Okay. That's more than I was expecting. Okay. So this is Antarctica and Greenland. Everything in blue is a mostly cloud-free stereo pair acquired in the last four years.
04:04
Everything in red is what I have received so far. And just to put it in – we're talking about 10,000 of these pairs in Greenland and roughly 20,000 in Antarctica. So huge amounts of data here. And just for context, if you go to mapfight.com and you've pit Antarctica versus the United
04:24
States, Antarctica is one and a half times the size of the United States, including Alaska. So to put this – we're talking about the continental U.S. plus. So clearly, GUI-based techniques to process these data are not going to work here when
04:40
we have this volume. And so what we've done is we've partnered up with some folks at the NASA Ames Research Center, the Intelligent Robotics Group there. And they have an existing pipeline called the Ames Stereo Pipeline. It's a set of automated open source command line tools. This was originally developed for a Mars Pathfinder back in 1997. And then in about 2008, they adopted it to process pretty much any image or taken
05:06
with any camera NASA has flown in the solar system. So up here, we have data from the Apollo mission, the Apollo metric camera. And this is a mosaic I put together for – this is Gale Crater on Mars. This is the Mars Science Laboratory Curiosity landing site here.
05:23
So in 2012, we received funding from the NASA Cryosphere Program to adapt these tools to process these Earth data. So at this point, we support the Digital Globe Rigorous Camera Model and any camera model that has RPCs. So these are rational polynomial coefficients.
05:40
Digital Globe distributes them. The Airbus, the Pleiades data, they have the RPCs. Skybox is doing them. And I believe Planet Labs has RPCs as well. Frank is sort of nodding. Okay. So all of these different images can be processed with these tools here. The core is written in C++.
06:01
It's mostly based on a custom computer vision library that's built to handle huge images in a very efficient way. Most of the core tools are multithreaded. They're memory efficient. And these tools are built to run on clusters. And so fortunately, I have quite a bit of time on the NASA Pleiades cluster, which
06:21
has enabled me to really do things that would take months or even years on just local workstations or smaller clusters. You can download binaries of this software package right now for Linux and OS X. And the source is always available on GitHub. And I'll give you links later. So in terms of how it actually works, it can be very simple.
06:41
There are a series of Python wrappers. The first line up there, stereo image one, image two, and an output directory. That'll run the entire pipeline and spit out a point cloud for you. And the second line is essentially a gridding utility that takes that point cloud and turns it into an elevation, a DEM, with syntax very similar to the GDAL utilities.
07:02
Or you can get in there and tune as many parameters as you'd like. And so it's highly customizable, which is great, because it's good and bad, as we all know. So in terms of how the pipeline actually works, up there on the left, the data are delivered in smaller tiles.
07:21
First thing we do is correct the original images to deal with some slight, not errors, but just shortcomings in the digital globe camera model. And then we mosaic them together. We do an orthorectification. So there are utilities here for fast orthorectification, multi-threaded orthorectification.
07:42
And then we generate a series of disparity maps to match features between the two images. We refine them, we filter them, output a point cloud, and then we output DEM and final orthorectified products. Most people really care about these things. So we're talking about a half an hour to 24 hours on a single dedicated workstation.
08:04
And that range is due to the fact that there's a lot of different settings. If you want a low quality product, it'll take you half an hour. If you want the highest quality product possible, you're looking at somewhere in the range of four to 12 hours, maybe even 24 hours. It also depends on the actual image data. If you have a lot of texture and features in your image,
08:23
it's going to run much faster than it would otherwise. In terms of the output products, we're posting these at about two meters per pixel. In terms of accuracy, our uncorrected vertical and horizontal accuracy, this is just out of the can from Digital Globe, about five meters vertical and five meters horizontal.
08:41
However, I spent a lot of time in the past few years bringing that number down. And I can say with confidence that if we have good control data, we can get below to sub-meter horizontal and vertical. Fortunately for the polar regions, we do have a lot of good control data. So NASA every summer is flying multiple aircraft to each pole.
09:00
This is a really cool mission. It's called Operation IceBridge. And they have multiple lidar instruments that are providing a lot of data, like those orange lines in the upper plots, with 10-centimeter accuracy. There's also existing data from laser altimeters, the ICESat mission, and then we have GPS campaigns. So we have a lot of points that I trust within 10 centimeters.
09:22
One of the recent additions to the Ames Stereo Pipeline is a utility to do automated point cloud to point cloud co-registration. So basically, this is an iterative closest point alignment, where you have one point cloud here. You have your other one that's close but not exactly aligned. And you can do a rigid body translation and rotation to fit the two to minimize the errors.
09:42
And so on the left there, what we have is the gridded DEM product that's uncorrected. The middle panel, I've clipped out lidar data over bedrock for this particular location. And we run this utility, and the output median error is roughly 45 centimeters for all of those locations after a three-and-a-half-meter shift.
10:01
There's a lot here, and I want to keep moving. But basically, the CE90 and LE90 numbers over there, those are the numbers that I'm getting for 31 DEMs over the same piece of bedrock in Greenland. So that's roughly what DigitalGlobe's spec is. So that's good. I believe them.
10:21
And this plot here is basically showing. So each blue and red pair is the before and after sampling of all of my control points for each DEM. And so you can see that before I do the correction, I have errors of four to five meters. And then afterwards, everything is down here about submeter.
10:43
And just to belabor this point even more, basically, this is that same stack. So I've got 31 DEMs covering this one piece of rock. And on the right here is the standard deviation of elevation values. And so some of these red areas are places where you have snow accumulating and melting.
11:02
There's some lakes and things. But these areas in here, that's just bedrock. And we can see that we have roughly, I don't know, 30 to 40 centimeter standard deviation there. So we really are submeter. And what this does is if I trust these elevations to 30 to 40 centimeters, then I can trust these elevations over the ice with the same kind of uncertainty.
11:24
Okay. So it's time for the pretty pictures. So this is Greenland. This is the west margin of the Greenland ice sheet. This is a mosaic that is 220 by 240 kilometers. And where these two dashed lines meet, that's the terminus of Jacobshaben Isbrae.
11:41
It's the fastest glacier on the planet. It's moving about 18 kilometers a year, which that's two meters an hour. So it's not typically what we think about when we think of glaciers. And it's dumping tremendous amounts of ice into the ocean. And actually, this is a fun fact. This is believed to be the glacier that calved the iceberg that sunk the Titanic.
12:02
So it produces these huge kilometer-sized icebergs. And it's changed a lot in the past few decades. So we have annual coverage here. And really, what's exciting to me is we have incredibly dense time series for the terminus, the very calving front.
12:21
So I have 46 of these Worldview DEMs in this five-year time period from 2008 to 2013. And I've supplemented with Tandem-X DEMs. This is the German X-band SAR mission, also beautiful products. But basically, we've got 57 of these DEMs. And we can bring them together and start to do some analyses here.
12:44
And there's a lot on this slide. But on the left, what I'm showing is basically a linear elevation trend at every single pixel at this location. And so what you see is that over the rock, the elevation trend is basically zero. So that's good. The rock is not moving.
13:01
But as we start to move up onto the ice sheet, our elevation values are dropping about 48 meters over the year. It's because the ice sheet is out of balance. There's more melting going on here that's being replaced by all the ice flooring here. But as we move even further up here, as we move out onto this fast-flowing trunk, this is the main glacier.
13:20
You see values that are somewhere 20 to 30 meters a year. That's a lot of elevation loss. In this five-year time period, we're talking about 100 to 150 meters of loss at that location. And on top of that, what we see is a very strong seasonal cycle. And I'm going to put an example of five points along that trunk. And you see that surface elevations are going up and down 30 to 50 meters seasonally.
13:43
This had never really been observed before. So pretty cool. I want to move on. This is a movie that I threw together showing what's happened in the past five years at this glacier. And these are preliminary products. And it's way too fast. And you guys aren't going to take anything away from this.
14:01
But I put it in here just because it's really cool. But also, we're effectively able in near real time to observe what these glaciers are doing. And that's unprecedented. We've never been able to do this before. Okay. I want to switch polls on you now. This is Antarctica.
14:20
How many people saw the press release maybe two months ago about unstable collapse of the West Antarctic Ice Sheet? Not in parentheses. It's happening. Okay. So a few of you. Well, that's happening right here. This is the part of West Antarctica that everybody's worried about because it's changing very rapidly.
14:41
And it's in a situation where there's really nothing that's going to stop a lot of the loss that we're seeing there right now. So fortunately, we have some excellent coverage here from the past four years. This is a mosaic I put together. This is 520 by 690 kilometers. So that's roughly the size of New Mexico to put this in scale.
15:02
And that's at two meters per pixel. And for reference, this is the best available data. This bedmap 2 DEM was produced about two years ago. And it's good. It's one to five kilometers a pixel. But when you compare it to something like this, there's some significant errors, hundreds of meters.
15:21
So this is a big change for us. And these new maps are basically enabling us to produce products like this. And what I'm showing here, this is now relative to sea level. So zero is roughly sea level. And all of this ice here is grounded. So this is ice that's in contact with the bedlock on the bottom.
15:41
Whereas all of this stuff, this is actually floating ice. That's an ice shelf where the ice is in contact with the bed and it comes out and meets the ocean and starts to float. And you see these beautiful melt channels. These are actually channels that are on the bottom of the ice shelf. The way we can identify those is because all of this floating ice is in hydrostatic equilibrium.
16:03
It's like an ice cube in your glass. So we're only seeing basically the upper 10% of the whole column thickness. So we're measuring this. But we can infer this because we know the density of ice. And so what this allows us to do is over time, again, we have data here for every year, we can basically track how the ice is thinning
16:24
but also how the ice shelves are melting over time. And this is just a preliminary product showing some of the elevation change maps for this location. This is the Pine Island Glacier. And over here, this is the previously available data. These are NASA ICESat satellite altimetry tracks.
16:41
And you can see this scale here, that's 50 kilometers. The spacing between a lot of these tracks is 10 to 30, 40 kilometers. And so we kind of had an idea that there was thinning going on here and elevation loss. But now we can actually pinpoint exactly where that's occurring, when it's occurring, and how much is occurring. So very exciting.
17:01
And stay tuned. Yeah, I'm at 1640. I'll briefly mention, these products are high enough resolution that we can actually use the actual DEM data to derive surface velocities. So this is essentially running feature tracking. We're using an image correlator to track every single pixel in one DEM at time one
17:23
to its new location in time two. And what that gives us is a complete three-dimensional vector for every single pixel from time one to time two and then time three, time four. And this is incredibly valuable from a glaciological standpoint. So very excited about this. And it's enabling us to do all kinds of new analyses.
17:42
And I don't want to get into the details here because I want to show you guys this. So I live in Seattle now. And I spend a lot of time out in these mountains in the Cascades. It's a beautiful place to go hiking. And our office, we have sort of a view of the top of Mount Rainier.
18:02
And I'm sitting there two years ago and saying to myself, why are we not collecting these data for the mountains here and the ice in Washington? Washington State has the highest concentration of ice of any state in the lower 48. Alaska blows Washington out of the water. But basically anywhere you see these blue polygons is where there's a glacier.
18:25
And each one of those red dots is where we put in a target to acquire these stereo data. And you can see these postage stamps. Those are locations where we got them in 2012 and 2013. So we did get some usable data. Unfortunately, a lot of our data looks like that.
18:41
This is the Pacific Northwest. And it's cloudy here. But we did get some good data for St. Helens. So these are some early products that I put together for St. Helens for 2012 and 2013. This is the Bare Earth LiDAR data for St. Helens. This is a really nice data set from 2003 back before LiDAR was really very prevalent.
19:07
And this is the WorldView stereo product. And I'll go back and forth just so you guys can kind of see. We're doing pretty well in terms of resolution and also in terms of alignment and accuracy.
19:21
And a lot of these no data gaps here, those are actually trees where the correlation failed. The more recent versions of the pipeline actually fill in a lot of those. But I didn't have a chance to make new slides. And just to kind of zoom in to really hammer this home, on the left is the LiDAR. On the right is the WorldView stereo DEM.
19:40
And the quality in terms of the resolution, some of these channels on the south side, we're able to resolve things that are several meters across. And if you look at these, you can see that things are changing there. Mount St. Helens is one of the most dynamic places on the planet right now. And there's this huge chunk of rock that came out of the ground.
20:03
And also a glacier has been forming there. So we can look at the elevation differences. And there's roughly 200 meters of new rock that erupted during the 2004 to 2008 eruption there. And the glacier, which there were two of them here.
20:21
And they've coalesced now. And they're flowing out the north side of the crater. So this is kind of one of the few success stories for glaciers in the world that we live in right now. I think it's really cool. So if we step back a little bit and saturate those changes,
20:40
we begin to see all kinds of other things. So when I first made this map, I saw these horrible colors around the outside. And I said, oh, I have a problem. But then I realized that this 2003 LiDAR data, it's bare earth data. So they've removed all of the canopy, all the vegetation. And with our stereo images, we're actually correlating on the tops of the trees.
21:02
And so this is a measurement of vegetation height, which for me is noise. I'm not interested in that. But for a lot of people, that's a signal. They're very interested in vegetation height over time, especially here, where you lost most of your forest in 1980. But it's kind of saturated.
21:21
You can't really see it. But these things, that's snow patch increases within valleys on the sides of the volcano. And then if you look closely up there at the Tugo River, there's actually a big signal there. There's been over 20 meters of erosion along this river in the past 10 years. And actually a lot of deposition up in the left corner there where you have these meanders.
21:43
And then that's a small landslide up there just below the Johnson Ridge Observatory. So this is just kind of a demonstration of what's possible with these data. And we're using them to make quantitative measurements of volume changes of ice to measure glacial advance and then to measure snowpack,
22:01
both winter accumulation and winter snowmelt, or summer snowmelt. And just a progress update, we've got 15 targets now. And the 2014 acquisitions have been going very well. We're getting data from April and May, and then again in September and October, as well as monthly stereo at Rainier. And does anybody recognize that?
22:26
Hood, that's Mount Hood. So this is everything that I've probably seen. I'm waiting for another delivery. But the data quality at Mount Hood is excellent. I don't have any results for changes yet because this is the only data that I have.
22:43
But we're in there somewhere. And there's Mount Hood. And just to give you a sense of the quality of these products, you know, I just threw it in here for the gratuitous 3D rendering. And yeah. So to summarize, we've got this automated pipeline.
23:00
It's working very well. We're producing two meter per pixel elevation models with five meter uncorrected accuracy and sub meter corrected accuracy with really, you know, I can do this with just nine ground control points. It doesn't take much. The way I see it, these repeat DEMs offer unprecedented spatial and temporal resolution
23:21
to study dynamic processes. And I see it, each of these is kind of like an airborne LIDAR survey. It's not going to replace LIDAR. But it's becoming close. And I think we can learn a lot, especially over rock and ice, with these data. And also, when you have these dense time series, if I can get these data every week,
23:42
it's an awful lot like having GPS receivers every two meters. And it enables you to do a lot of things that you wouldn't be able to do otherwise. We're generating mosaics. And we're going to be generating mosaics that are over a million square kilometers. And the baselines for these are established at this point.
24:00
And we're just starting this program. And I hope that we're going to continue to be getting these data 10 or 20 years from now. So yeah, I'll skip that. And here's the info. And I'll stop there. I encourage you to download and check out this package. It's cool. There's some useful utilities in there. And I'm happy to take any questions now or after the session.
24:23
Thanks. Was that the ELWA that I saw up there as a target area?
24:43
Yeah, we got some data for the Olympics. Have you been making any products or using that at all? I haven't gotten these yet. My plan is to wait until things finish shooting in October.
25:05
And then batch order everything and run it through. I should mention, to batch process a data set like this on the Pleiades cluster, we're talking about half a day. Because I can just distribute to multiple nodes. And it's really pretty powerful.
25:24
About how big do the DEMs get once you're done processing them? In terms of file size? File size. It depends on the final output resolution. And they compress pretty well, usually. So for a postage stamp like this,
25:41
we're probably talking about a few hundred megabytes. But if we go back to some of these other ones, each of these strips, that's a degree of latitude. File sizes definitely start to increase. In some cases, they can be pretty big.
26:11
Have you thought about using this in forestry applications, such that you would have like a bare earth layer and then stand height or whatever? I've thought about it, but I've got enough to work on already.
26:23
I have started talking to some folks at Goddard who are really interested in those products. This is only the beginning. I'm sure that these will be used for forestry applications. Can you go back through the Digital Globe archive and use any pairs if they weren't flown as shots together?
26:45
Yeah, so like a coincident stereo pair. I can do that assuming that my surface isn't changing. So for places like the Cascades, that's usually I can take an image acquired on one day and an image acquired maybe a week later,
27:01
and I can probably generate a valid DEM that way. For somewhere like Jacobshavnizbre that's moving two meters an hour, I can't. My maximum time window there is probably about 6 to 12 hours. And I've tried. But I have actually made some stereo pairs from non-targeted along track observations
27:21
and they've worked very well. Thank you everybody.