Merken
Point Clouds in a Browser with WebGL
Automatisierte Medienanalyse
Diese automatischen Videoanalysen setzt das TIB|AV-Portal ein:
Szenenerkennung — Shot Boundary Detection segmentiert das Video anhand von Bildmerkmalen. Ein daraus erzeugtes visuelles Inhaltsverzeichnis gibt einen schnellen Überblick über den Inhalt des Videos und bietet einen zielgenauen Zugriff.
Texterkennung – Intelligent Character Recognition erfasst, indexiert und macht geschriebene Sprache (zum Beispiel Text auf Folien) durchsuchbar.
Spracherkennung – Speech to Text notiert die gesprochene Sprache im Video in Form eines Transkripts, das durchsuchbar ist.
Bilderkennung – Visual Concept Detection indexiert das Bewegtbild mit fachspezifischen und fächerübergreifenden visuellen Konzepten (zum Beispiel Landschaft, Fassadendetail, technische Zeichnung, Computeranimation oder Vorlesung).
Verschlagwortung – Named Entity Recognition beschreibt die einzelnen Videosegmente mit semantisch verknüpften Sachbegriffen. Synonyme oder Unterbegriffe von eingegebenen Suchbegriffen können dadurch automatisch mitgesucht werden, was die Treffermenge erweitert.
Erkannte Entitäten
Sprachtranskript
00:08
I'm there way run so I want to
00:13
have my my topic is about 2 point clouds and you the how to show them the browser of his project called poetry and the presentation was prepared by by me and my question the original author of this library was unfortunate label to come into force for the 2
00:32
and about the that few more words about me and you grow from members of the developers and maintainers of a project called think founded the company Europe public and I'm living in Germany in summer and the rest of the year in Japan and I for long time I enjoy open-source hostility and I'm always and and you know it happened that I have to work was point cloud data as a project and there so what do we do with
01:01
this point cloud data and maybe point out that there not not so common so uh it's a
01:09
point cloud data recorded and to maintain them by road infrastructure so it's difficult to pee looks like this and it's very high quality and live how little details that along the road and there's just no data and with moral and you very complex junctions so sometimes it's very dense and sometimes there's nothing
01:33
and so that uh what you need to collect the status of such it's quite expensive and know that something you you you buy so easy but the other companies
01:46
that use point the laser scanners and all these things and it's getting to the point of the is getting more and more and also new cars have
01:58
tons of senders and also scan the environment so I think it's just a matter of time until you come across part of the text you can do that yet i and that you
02:11
can't wait there's also a way for you can create your own point of data and you can do that work and you can maybe
02:18
you have access to such a strong and can play around that while working and so you can go to the next part I can fly around a
02:26
bit and take photos from the top and if you look into your
02:31
computer model files and then there's a lot of open source software that can turn this into a point
02:40
and then I did this 1 times and I forgot to to adjust the lens of the camera and was officially service and became drawn and but I I found it very cool and was fun to play with and so the park in the end I looked at this just flying around for the system is up to the battery was off and so yeah led to my computer
03:01
running and it was running for quite some time and became very hot I couldn't work anymore profiles for the times and it turned out that this is a small part of it was already 45 million points and the to 1 . 2 D about last for 1 and so you welcome to big data and so
03:20
on the question is now if I want to show this how can we showed but millions of billions of points so that the browser and that was the requirement for the project we did and so I was looking around and what existed was wasn't to was 3 years ago and
03:38
I found a protocol poetry and that was the very interesting and it was uh I I mean that way you point cloud your for larger data sets and it was that have developed would still developed by should from Vienna and the number of contributors growing since then and so know maybe will the 1st during public was the 1st of companies supporting this financially and and then rapidly so that contributions for new funding new features and more and more came so the library was actually growing in it can of open source project and no I I didn't know exactly when it started and the creative community started out at the top from started in February 2000 14 and there was a lot of activity depending on funding and then there was a break in the last month or so take a breath and I think we have more in mind and development will take increase again and there's a whole for works so there are few things when you have such as a lot of a lot of data to deal with usually you only can see a small part of the data you don't have to show 45 million or more points you don't have to load everything so only what you can see the point of from the point of view this is original so that it uses a multiresolution tree and this allows you to their not displayed and everything that is unnecessary and only need 1 Lord what you need so you try to minimize the number of points and so it has a lower level nodes that contain the resolution models of large areas so when you zoom out your and again this and when you zoom in on they resolution increases and the coverage the area becomes less when assuming and this is pretty much what we have also this not tied so and that's it's a very common 1 way to to deal with more such amount of data and each octree notice stored in separate files and so the cell is only necessary of 5 and in our case for example we store the uh the files on street and as we have kind of unlimited storage and there's also a certification requirements to 2 nodes to Lotus and it's so poetry
06:21
requires certain data format and for that you know that we needed to it so there's a converted that there was a need for this and then departure converter and In the initially it was not the case but then we we decided to use 3 JS as a base for the rendering engine and so you have all the capabilities that you would also get when you use Regius so if you want to extend this functionality short read your inside of so you can do and there was also added client-side last and the compressed version of last support from project proposal which is also very cool and I'm not really needed some tools especially for measurement and the point clouds which was added and health costs open source and it's a very business-friendly license to use so what the data format that is currently four-port supported you can take as input lasts compressed version and PY files and as an output of the run the converter you get to the colleges file which is a metadata file that contains information about ongoing spacing storage format and so on and you get a menu of tree nodes and and they contain information that position cholera intensity and justification information and there's an additional file and to store the octree hierarchy so the features 1 transistors and I will show some some interesting ones later in detail so 1st you can read the billions of points in the process that was that was the main goal and you can change the point size so if you have more sparse point cloud you can feel that fill the holes and uh there's something called I don't lighting that makes the point loads of shiny and that in a minute and you can render RGB elevation intensity or just about any attributes that that you have and there are various rendering modes so if you powerful hardware that you can do more that things like interpolation or spreads and if if you look at the computers also strongly condemned decrease this that and computation intensive interpolation and remote and so you can measure distance and areas where you can get a high profiles and you can click volumes and a lot of this functionality was added on request and funded projects that needed it so
09:05
the whole input you show a point clouds more nice nicely than just the points and that was added later and what makes it a very nice is this interpolation of the 2 interpolation and it increases the readability of teachers like fonts and become part of the easier to to to read and you avoid overlapping points that by using some their nearest neighbor interpolation or or whatever the want way by all of the points to get the and the other thing to
09:40
make it look more nice it's called vital lighting so you that some some illumination and at the same time you some all kinds and so that it feels more like most 3 D and more depth and and so it looks quite nice then and this is especially useful if you rent elevation or classification of few few data and some something else that
10:09
was added to and sometimes it's very difficult to navigate in your in your this area and 3 D so you have some some points of interest and you you can annotations numbers small that you would like POI somewhat and when you click when you click on them and it will resume to that point and then you can you can pre-defined uh call for you look at this at this point for this object and from something that
10:43
that we needed especially was and getting height profiles and initially we kept it really didn't calculated directly integrals about that data is already there so you don't have that just by you draw a line that can also call consist of multiple segments and from there it will immediately draw you hide profile like you see on the right side and it will do show it even if you are few lines is exactly to show it and uh flat into and because I can't and that there was come to find such a nice path so here's a nice example that I'm not sure how well it works with the Internet and in the most and then there is something that the
11:30
measurement that was important for us but the more you see here there's an example panels but of course is a library so you can use the functions and methods to build this yourself but you can get an idea what what you could do so that we can display coordinates of like at a certain point and you can't the actual distances like uh the middle picture you can I removed like same if you measure something and new move it again and you can also show this department and you can show that 2 D profile for this slide the
12:16
that's would try to show they want it and in case it doesn't work I will show you the pictures quickly so you see on the left side this it was initially for debugging but it's also a good point to start with so you can make a lot of adjustments and to your pointed out you can change the point size you can show how the
12:33
maximum amount of points you want to you want to render them because somebody if your computer is not very powerful and this uses a lot of of computation power and so you know it's good to to to reduce the maximum amount of points to display and you can do things like that change and the point size to it like automatic or a fixed size and um the the shape of the points and all the tools of our artists of their or if you
13:06
zoom in and with the height with the coloring that it actually it's quite beautiful and there are actually many many
13:17
examples and data that that was allowed to so Marcus was able to get this data and show it in his the most and all that they must also the source code is available on and there are there are
13:34
some limitations on the promoter mutations 3 but I wanted to mention these 3 so 1 is the the browser security policies so even if you could run it from a file on there so you have to you have to have a web server to to to load to note this stuff that human transcript released in the works of uh and um of course you need uh virtually video capital browsers and devices and when this is becoming so that if you want to use ago all this was a big issue but in the meanwhile this that's a big problem anymore and also Internet Explorer knowledge eventually what I think is working with a little bit and then you can only use it for viewing and analyzing you can manipulate the data course because it's stored in the file system and so who who is using quadtree under there are quite a lot of companies I heard by using it maybe they're more probably them more and we just don't know about it and so yeah it's I think it's quite as activity is so well use projects and and also many many fragment much function was funded so this seems to be quite some demand but they are likely like friends projects and very interesting ones and uh some I would just quickly mention and later they will be for example a talk about the items project of this which is a 3 D virtual geo-spatial Uses dieselization project and then in the afternoon there will be another session and then we can use picking him what find and that does indexing of point clouds and it's something that might replace the con the poultry converted eventually so at it does it in a very sophisticated way and then this greyhounds that provides data streaming in surrounding environment and poodle speed of you will also hear more probably in the afternoon and I mentioned before that so treaty titles and something we are also looking at and I'm not sure if we want this state of this is 1 that I think we are missing some some standards and and as the PG point cloud project that was maybe 1 of the 1st ones and but I haven't looked at it I haven't tried it and much and I think the point clouds a topic that many people care about these days and like
16:07
I said before I think we need a standard the standard is what what makes an open-source stronger so when we can easily exchange libraries or and we can easily communicate with other projects and so but I think it's important to to think about and and define them and maybe we can't wait to OGC finished their their official stand that and none other that's so
16:34
quickly show try to demo and
16:38
so when you this is that the data sets that has 150 million points showed before the other side you can turn on and off and this the configuration menu and for example when I click on points to at most to this point
17:00
and is it loads maybe the Internet speed is not so high this that's not to make the points adaptive it is also that the
17:11
graphics card that's the screenshot from from before it can decrease the points where you can increase so just this compute powerful In this case we set the maximum to 5 million points and and the default 1 was 1 . 5 million points you can change opacity you can this is from 3 days even at the sky and I like the background and for for
17:45
measurement for example it's but the problem with the most so this is still high they can control line that try to hit the home that's a good and bad and now we can show that the profile here on the bottom which is not so good like the 1 in the example and you can you can show at which point at which point you are so this is a this is for for them the other application perspective the standard interface that shows what what country what we can do and
18:49
and so after some some few months of break of development knowledge it's time so there work on making the so what do readers candidate of version 1 . 4 was released and that should be on the final release should be available by the end of the year and that's what 1 of the new features or will if you do faster convergence of the data so when you convert your data and from from what has to be has been improved like there were too many files and on so you have to find the right balance between the amount of files and how you all for your body organized in folders companies and in this case we use 50 per cent less space now and has fewer fires and should improve the performance and it should also contain support for a 120 which was implemented but how about life the future exactly who was involved and that there later other ideas like predicting maps of 2 point clouds and in real time or and load and display . com for months without any conversion and then there is so that is always needs funding
20:08
so if somebody needs that have never actually get that far because of the support of many companies and so I think there is a lot more possible and if you are 1 of those who
20:18
need something that and the contact me or content models and then you can find the project on www . treated work there and we are very welcome to hear the bad things that work and the things that don't work me and you're emerge and becomes because few that you will much the sole questions like ensure you have more questions than you can that have come and take the possibility cool export to initiate could be Prof and eat into each form to that I think you can do this and yeah it's it's rendered as their peers Street Street Suite of curious but I think there is a way to export exporters but I'm there and I don't know how it works maybe I think it is possible because you may be briefly explained that the tool chain so assuming I have up on clouds which where so I'm from the German railways we have actual trains which ability their skin and we have point clouds assume I want to make this kind of prototype of this kind of them onto each due to all management what you will how how that perceived how would I
21:47
know which tool chains always today use you have
21:51
your your last files and those guys I assume so and then you and then use but she converted to to to converge in the form of that is but look from 1 and then you have to approach to terrestrial and then if new data arrives so you can actually you can you don't have to do do everything from scratch but there but I think there's going to be more sophisticated variables and the the focus of poetry is not so much to convert soul and this and and and Greyhound streaming service for this may be something that scales for for larger areas so in our case we always have some conjunctions and certain areas and they're connected and we get the data once and then we run it once through this converter
22:42
yeah this wonderful measurements is it possible to capture the data and exported to other programs and so there's
22:50
nothing building like this no the it should be possible to to do that you can also you can also I changed the rates of your of the slice of the profile like the image of what you want to have the point itself or just an image what what kind of for much of the because you only you don't show it don't have all points of your original data now I was thinking about it when
23:16
you have the missionary anyway hearing it you wish I how you could measure and digitizes some sort of digitizing but is it possible to capture this information back into another prominent the provide but I think that the country itself is
23:34
and so you you don't keep all the points you on the show you only show you only know what what you need so the data you have in your browser maybe it should be possible to add this functionality to true to exported but I think maybe image makes more more sense like the question before another question the order the for I'm a nice presentation just a comment field also has such a possibility tool and create a little Donald portal for the original data so if you
24:20
in addition to the last input you give projection it will create OpenStreetMap which you can download still alias tiles took just create a very simple model called of OK that that that you have this so
24:36
we don't use the slide and all that exists what is this seems to be some something done in this way thank you for attending 1 last question maybe and OK thank you Damian for
24:57
business with a vision fj
00:00
Umsetzung <Informatik>
Virtualisierung
Extrempunkt
Browser
Datensichtgerät
Program Slicing
Twitter <Softwareplattform>
Information
Dicke
Raum-Zeit
Computeranimation
Netzwerktopologie
Metadaten
Last
Dämpfung
Softwarewartung
Computersicherheit
BCH-Code
Kontrollstruktur
Dateiverwaltung
Punkt
Oktalbaum
Gerade
Cliquenweite
Schnittstelle
Automatische Indexierung
Addition
Suite <Programmpaket>
Shape <Informatik>
Sichtenkonzept
Hardware
Computersicherheit
Gebäude <Mathematik>
Globale Beleuchtung
Profil <Aerodynamik>
Schwach besetzte Matrix
Bitrate
Software
Dienst <Informatik>
Knotenpunkt
Menge
Forcing
Interpolation
Zahlenbereich
Ein-Ausgabe
Client
Server
Messprozess
Routing
Programmbibliothek
Ordnung <Mathematik>
Programmierumgebung
Multiplikation
Wellenpaket
Mathematisierung
Maßerweiterung
Datensichtgerät
Open Source
Mailing-Liste
Variable
Bildschirmmaske
Informationsmodellierung
Knotenmenge
Oktalbaum
Perspektive
Programmbibliothek
Skript <Programm>
Abstand
Spezifisches Volumen
Konfigurationsraum
Attributierte Grammatik
Videospiel
Protokoll <Datenverarbeitungssystem>
Raum-Zeit
Open Source
Browser
Peer-to-Peer-Netz
Elektronische Publikation
Benutzerprofil
Netzwerktopologie
Summengleichung
Abstand
Echtzeitsystem
Parkettierung
Fibonacci-Folge
Wort <Informatik>
Interpolation
Streuungsdiagramm
Viewer
Bit
Punkt
Prozess <Physik>
Versionsverwaltung
Kartesische Koordinaten
Computerunterstütztes Verfahren
Computer
Analysis
Ähnlichkeitsgeometrie
Übergang
Eins
Videokonferenz
Internetworking
Mooresches Gesetz
Font
Digitalsignal
Datenmanagement
Maßstab
RPC
Font
Parkettierung
Minimum
Volumen
Volumenvisualisierung
Flächeninhalt
Maschinelles Sehen
Default
Einflussgröße
Prototyping
Bildauflösung
Funktion <Mathematik>
Zentrische Streckung
ATM
Lineares Funktional
Softwareentwickler
Teilbarkeit
Volumenvisualisierung
Speicher <Informatik>
Applet
Quellencodierung
Systemaufruf
Übergang
Programmierumgebung
Quellcode
Ein-Ausgabe
Knotenmenge
Dateiformat
Digitale Photographie
Softwarewartung
Rechenschieber
Hierarchische Struktur
Datenfeld
Verkettung <Informatik>
Funktion <Mathematik>
Automatische Indexierung
ATM
Dateiformat
Projektive Ebene
Information
Versionsverwaltung
Aggregatzustand
Standardabweichung
Server
Belegleser
Metadaten
Ortsoperator
Besprechung/Interview
Zahlenbereich
Zellularer Automat
Hierarchische Struktur
E-Mail
Kombinatorische Gruppentheorie
Graphikkarte
Physikalisches System
Graph
Benutzerbeteiligung
Digitale Photographie
Software
COM
Inverser Limes
Optimierung
Speicher <Informatik>
Softwareentwickler
Bildgebendes Verfahren
Computersimulation
Streuungsdiagramm
Einfach zusammenhängender Raum
Autorisierung
Elektronische Publikation
Digitales Zertifikat
Bildauflösung
Methode der kleinsten Quadrate
Physikalisches System
Umsetzung <Informatik>
Sichtenkonzept
Fokalpunkt
Quick-Sort
Objekt <Kategorie>
Mapping <Computergraphik>
Flächeninhalt
Last
Mereologie
Surjektivität
Visualisierung
Modelltheorie
Metadaten
Formale Metadaten
| Titel | Point Clouds in a Browser with WebGL |
| Serientitel | FOSS4G Bonn 2016 |
| Teil | 146 |
| Anzahl der Teile | 193 |
| Autor |
Kastl, Daniel (Georepublic)
|
| 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 | 10.5446/20360 |
| Herausgeber | FOSS4G, Open Source Geospatial Foundation (OSGeo) |
| Erscheinungsjahr | 2016 |
| Sprache | Englisch |
Inhaltliche Metadaten
| Fachgebiet | Informatik |
| Abstract | Potree is an open source project that implements point cloud rendering capability in a browser. It is a WebGL based point cloud viewer for large datasets. Thanks to WebGL, it runs in all major browsers without plugins. Over the past years Potree has evolved from a small library to an active open source project with an active community, several companies funding development and an increasing user base. This presentation will give an overview over the current state of point cloud rendering with Potree, about the difficulties and challenges. Pointcloud data is expected to play an increasing role in the next years with falling prices for previously very expensive hardware such as laser scanners, the development of autonomous vehicles and the popularity of drones. Powerful hardware and WebGL will open up a wide range of innovative browser-based web services in the near future. |
Zugehöriges Material
Ähnliche Filme
