Collecting data with UAV - Using data in QGIS
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 | 37 | |
| Autor | ||
| Lizenz | CC-Namensnennung - Weitergabe unter gleichen Bedingungen 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 und das Werk bzw. diesen Inhalt auch in veränderter Form nur unter den Bedingungen dieser Lizenz weitergeben. | |
| Identifikatoren | 10.5446/40771 (DOI) | |
| Herausgeber | ||
| Erscheinungsjahr | ||
| Sprache |
Inhaltliche Metadaten
| Fachgebiet | ||
| Genre | ||
| Schlagwörter |
QGIS ACoruña Konferenz 201928 / 37
3
5
8
13
17
19
26
00:00
SoftwareentwicklerKopenhagener DeutungCoxeter-GruppeProzess <Informatik>SpeicherverwaltungComputeranimation
00:26
GrenzschichtablösungProjektive Ebenet-TestDifferenteDigitale PhotographieMatrizenrechnungTwitter <Softwareplattform>RechenwerkEinsGraphfärbungKlassenkörpertheorieQuick-SortTaskBitRechter WinkelVarietät <Mathematik>ComputerspielZahlenbereichVorzeichen <Mathematik>ProgrammierungTopologie
02:34
SpezialrechnerTypentheorieKopenhagener DeutungPunktPunktwolkePunktwolkeBildgebendes VerfahrenRelativitätstheorieLesen <Datenverarbeitung>GrundraumPunktComputerspielProfil <Aerodynamik>
03:11
PunktPunktwolkeKopenhagener DeutungSpezialrechnerProzess <Informatik>SichtenkonzeptRechenbuchDemo <Programm>TopologieViewerWasserdampftafelPunktAutomatische IndexierungOffene MengeBildgebendes VerfahrenFlächentheorieMomentenproblemCliquenweitePunktwolkeCoxeter-GruppeEndliche ModelltheorieStreaming <Kommunikationstechnik>Profil <Aerodynamik>PolygonnetzDifferenteSpezifisches VolumenPixelProdukt <Mathematik>ResultanteRechenbuchBitmap-GraphikGebäude <Mathematik>DialektDienst <Informatik>Computerunterstützte ÜbersetzungTopologieWeb SiteFlächeninhaltWald <Graphentheorie>MultiplikationsoperatorRechter WinkelVollständiger VerbandArithmetische FolgeFunktionalStrategisches SpielKlasse <Mathematik>ClientMetropolitan area networkGesetz <Physik>Aggregatzustandt-TestComputeranimation
09:10
Kopenhagener DeutungPunktPunktwolkeService providerViewerDemo <Programm>Prozess <Informatik>Desintegration <Mathematik>TopologieSichtenkonzeptQuellcodeTermRoutingGeradeBildgebendes VerfahrenGüte der AnpassungPunktZahlenbereichResultanteBildschirmmaskeKontrast <Statistik>Prozess <Informatik>TVD-VerfahrenPartikelsystemMobiles InternetProgrammierungWort <Informatik>DialektSolitärspielSichtenkonzeptDemoszene <Programmierung>ComputerspielMetropolitan area networkAggregatzustandSoftwarewartungAutomatische IndexierungStellenringVorzeichen <Mathematik>VererbungshierarchieFunktionalMathematikSystemaufrufNatürliche ZahlMessage-PassingPunktwolkeDatenstrukturMereologieOpen SourceCoxeter-GruppeParametersystemFilter <Stochastik>Elektronische PublikationGraphfärbungDifferenteSoftwareentwicklerTaskStrömungsrichtungQuick-SortSoftwareViewerFlächeninhaltKeller <Informatik>Service providerComputeranimation
16:17
PunktPunktwolkeLeistung <Physik>Textur-MappingNeuroinformatikBildgebendes VerfahrenResultanteGebäude <Mathematik>InformationRechter WinkelUnrundheitFehlermeldungMetropolitan area networkWort <Informatik>MultiplikationsoperatorRechenwerkGraphfärbungMereologieComputeranimation
17:30
EDV-BeratungComputeranimationJSONXMLUML
Transkript: Englisch(automatisch erzeugt)
00:05
This is a presentation which we had only a few hours to start or do, so keep that in mind when we talk. We did a workshop two days ago about how to process drone data and how to use QTIS.
00:27
We have a lot of drones that during the last few years, for most people, they can afford to have a small drone or they need to have a large drone too. We started with this one here, a Phantom 4, and we did a lot of collecting data,
00:48
some were rendered and made all the photos. Then we were lucky to get the small Mavic, put it in our handbag and collecting even more data.
01:04
It's been fun both for our students but also for us because we found a lot of different data types. Now we were in Sanzibar, in Foshuaji, in September last year, and there we saw a huge project.
01:25
The entire Sanzibar was filmed with a UAV, several UAVs. And Kenita, I can't pronounce her name correctly, she was in charge of this project, and they were simply measuring, making all the photos from drones.
01:46
Here at the top right, the picture here, you can see our latest toy, it's a Matrice 600. We were lucky to get funded, and underneath here, you can see we don't only have a camera,
02:02
but now we also have a LiDAR. And if I can follow Twitter just a little bit, LiDAR is the new black. Everyone wants to have a LiDAR, it's the toy, not toy, but it's very exciting to get data from that kind of tool.
02:26
Earlier today we heard about people having underwater UAV to collect data too. So what kind of data can we retrieve?
02:41
Normally we see RGB images, we can also have spectral images, and we can have point clouds. And the point clouds here, which you're seeing here, is made from LiDAR.
03:00
It's the latest here, you can see from our forestry college. And you can both see it from above, but you can also see the profile, which was made here. When we are processing the images, we can get different, what's it called, products.
03:21
As I said before, we can have autofotos, where we have images which we are creating into autofotos, which we can use to digitize from, as a background image. But we can also produce vegetation indexes if we have these spectral cameras,
03:45
so it's going to be used in both forestry, but also in farming. We can produce meshes, you just heard a presentation about meshes, but if we take these aerial images, we also get meshes as a side product.
04:08
I haven't been using meshes yet, but after this presentation, where we heard about MDAL, we are going to look more into that, because I wasn't aware what to do with this kind of data.
04:26
But we also got the point cloud, point cloud from images. When I heard it the first time, I thought, this can't be true. You can't make a 3D model from an image. Of course, I've seen from the old days where I have two images,
04:44
and put them together, and then I could imagine this 3D, but making it 3D into a GIS, that was, no, it can't be true. But it is true, and from that we can make surface models. At our workshop, we made surface models, we made terrain models,
05:04
and we made terrain models where it was included with buildings, so we could make even more different products. So we were extracting raster data, and we were extracting pixel data.
05:27
When we have these point clouds, point clouds, we can't open point clouds in QGIS at the moment. We can't view them yet. We can view them, that is very primitive, if we want to view the point clouds in QGIS.
05:45
But we can classify the point clouds, we can extract different kinds of data, we can calculate point clouds, and we can also make some publishing so we can see the point clouds, but not in QGIS.
06:04
We can see it, for instance, in poetry. I think I would like to show you some of the results. This is the first one made from our first flight three weeks ago with our drone.
06:33
It's in the forestry college, and some of you have been there. And you can see these are the buildings here, and here are the climbing coals.
06:44
And if we look here, right here, we can see, it's too small at the moment, but we could see the person standing down here, but we can take a small walk in here. And this looks amazing.
07:02
Why go outside when I can sit in the door? And if I zoom out, I forgot my mouse, so let's do this here. We can even make a profile. Oh, you can see right here,
07:23
can you remember, there was a high tower, and we could walk in here, it's right in here. Let me take you to this profile, make a profile, and make the profile a little bit wider.
07:51
Then I can show the profile in 2D, perhaps, coming here. So this is the kind of data we can get from a UAV.
08:05
This was the LiDAR, but I also have the same kind of data, of course not with the same resolution, not with the same amount of points, from aerial images. And it's amazing what we can calculate.
08:22
We can find single trees, we can measure the height of the single trees, we can measure the ground width and the volume of the ground. So in forestry, this is quite important stuff. If we classify the point clouds, we can remove all the buildings, remove all the trees,
08:46
and simply show just the ground, or choose to show the ground and the buildings, so we can calculate streams around, when it's raining a lot, how does the water runs around the streams here.
09:01
But we need something more. So we are very lucky, because the previous presentation, from Martin and Pete, also talked about these issues.
09:21
So, in our drone workshop, we are just using a very small point cloud, with just a few thousand of points, and we can use LAS tools, or we can use PDAL, as I did, I did all the processing, without processing outside the QGs.
09:47
But what we would like to have in QGs, besides GDAL, MDAL, is PDAL. PDAL, we are integrated, and we are looking forward to the presentation about this.
10:02
But we would like to have a native provider to read these point clouds files in QGs. And also, we have another problem. These point clouds are really huge. This drone captures a compressed one gigabyte of points per minute.
10:27
So, after a few minutes, we will have a problem. So, we might also have some provider able to read it from something like Entwine,
10:40
or some streaming provider that we can ask for more points, and for more points. And we also need a point cloud viewer. And my question is for Martin. Are we able to use the current viewer for large amounts of points? Should we have, or integrate with something like Portree or PlusIO?
11:03
How do we do that? It's our question. So, basically, since we are using these large point clouds, we'd like to have it more integrated in QGs. And we think that images are right now very less expensive to have,
11:24
everybody has cameras, everybody can have drones. So, we will have more and more point clouds able to be processed, and we'd like to do that in QGs. And I think we will end it here.
11:42
Just a second thing. It's not about point clouds, that's about the images. I didn't tell about it, because it's already there for the spectral cameras. We can use a semi-classification plugin, and it's already there with good documentation.
12:02
So, we tried to do the itch for the point clouds, that's what we hope for the developers to look into. Very much.
12:29
What are you using for the step from the images to the point clouds, when you do that sort of structure from motion step? Yes. Please repeat the question.
12:40
What kind of programs do we use, from images to point clouds? At the moment, we're using different programs, I use Pix4D, and OpenDroneMap. Actually, OpenDroneMap was what we had this workshop.
13:01
OpenDroneMap uses OpenSFM, which is a non-implementation of structure from motion. And right now, we are using all of these tools outside QG. So, we are able to do the structure from motion in the common line, and even as the output, the orientation, the size of the image, and so on.
13:22
To classify and to compute things, we can use PDAL, and we can try to use different filters. We have a lot of different parameters to classify things, but it's all done outside QG, running on the common line,
13:40
which is not bad, but the user experience for some users is not so good, as something that I'm true. I'm able to select just the number two, which is for the terrain, and see just the terrain, and see just the buildings, and so on. So, we are using, and in this workshop, I think it went well,
14:04
but we're not able to see all the difference between the two software stacks, because we are trying to use different software stacks to compare the results, and to see how good our open source tools are against the other tools.
14:23
Secondly, the result of, for instance, the point clouds, we get an unclassified point cloud. They are pretty nice for both products, but when it comes to classifying the point clouds, none of them are really good yet, because in OpenDromeMap and in Pix2D,
14:41
you cannot say this is a wilderness nature, or this is an urban, or this is an urban with high buildings. That's not possible. For that, we actually use, I use last tools, because I can make those, I can choose those parameters
15:00
in this part of the program. And last tools does really work nice in cookies, but some of the functionality you have to pay for. The classification is free, yes, but when it comes to making the DEM, you have to pay.
15:27
This might be a naive question. I'm not familiar with the point clouds, but I've seen some point clouds with the imagery draped onto it, like you showed the point cloud, which is like discovered by eye, but can you trace the imagery with point clouds?
15:41
Yes. Yes, we can. This is almost simple for aerial drone. It's not so simple at all for mobile. The color of the point clouds, the images from above,
16:03
it's not a simple task. So, as you said, we have multiple colors, depending on the work of the work or the technology, so it's not a simple task to go. It's much more simple for the eye area.
16:20
When Ling is preparing for the drone flight, we usually use the camera pointing to the ground, 90 degrees, but sometimes to capture the textures and to render the buildings with the images, it's good to fly with the camera pointing, for example, 70 degrees,
16:42
so we will have the textures of the buildings and so on. We are able to join this information and put the images very, very small and put them on top of the point clouds. You can see it as a result of what's here, when we have more power on the computer.
17:01
But when you get the point cloud, you have the X, Y and Z coordinates, you have a classification in the beginning, it's unclassified and you have the RGB colors, so you can add that to the points. So it's very amazing what you can really achieve, so please come and see that.