Moving

Dear People,

I have stopped Blogging under this domain. But I have not stopped altogether.

Read on at: http://blog.technokrat.nl

Love and kisses,

Fritz

Whoops I FLARted..

Today I fooled around a bit with the FLARManager (tutorials graciously provided by transmote). I tried to get some Papervision 3D objects going using all these different tutorials I have been following over the last few weeks (e.g. papervision2.com and rawkes.com). But it has not been as easy as I had expected. But then again, things never are.

The difficulty is not so much in displaying 3D objects, because pv3d and flarmanager do all the thinking for you. But trying to get from a GIS file to the right formats and the right information to be displayed in the right way, is still a bit tricky. The model still has to correspond with the actual data that was used, but more often than not these models contain very dense data on a large scale. For now the model has been so big that I only saw a small fraction of it in my display the rest hovering quietly above it. Need to find a smarter way to doing this than:
3dobject.scale = 0.001;
But I’ll get there I’m sure.

For now I can render multiple cows on my desktop. And at least that is something right?

Property of Papervision2.com I think

P.S. Why cows instead of GIS (the model is just small and loads quickly, good for testing purposes.)

Thoughts on LiDAR

Yesterday I was fortunate enough to be able to visit the European LiDAR Mapping Forum (ELMF) in The Hague. For those who do not know what LiDAR is, it is a sensor that acquires high-quality (very dense) 3D data. If you would like to check out some of the data, what it looks like and play around with it a bit, here is a “get-started”. It is further explained on Wikipedia, to which you should give some money for using it on a daily basis, if you haven’t done so already. Back on-topic I believe it was quite a successful event, some 600 applicants joined in, which was above expectations I gathered from Alastair MacDonald‘s closing remarks.

Some interesting applications (to me)
Mandli’s examples of Mobile Mapping and quick high-quality 3D model creation, with the help of Mobile Mappers. Mobile Mappers are another example of a cool technology. These cars with LiDAR sensors mounted on them can collect 3D data at highway-speed. Awesome.

Topcon is collaborating with the company creating the Moto GP 10/11 game. This game uses a model of the Silverstone track gathered and created in about 2 hours using a Mobile Mapper. The conference was premiering some of the stunning 3D imagery created for the game.

Railtracks around Dover. To monitor land slides and land changes at the cliffs of Dover they showcased quarterly surveys of the area using Terrestial Laser Scanners.

Beyond understanding
Most of the rest of the talks were less practical of nature. This is not bad in any case. On the contrary I found these lectures very interesting. However for my studies and thesis it will probably not be of any importance. Although I now (claim to) understand why you should prefer Discrete Return over Full Waveform for most purposes.
Other things I just didn’t understand were all over the place. But it was very educational and interesting nonetheless.

One more example before I leave you. The Dutch branch of TerraImaging also had a stand there. They deliver AHN2 images (the Actual Elevation of the Netherlands version 2) data. Awesome detail. With an accuracy of 5cm they deliver Elevation Maps of areas in the Netherlands. Just don’t think that data comes without cost. There is a free online viewer though at the AHN website.

Rendered Useless

A little update on what I am doing. Today is a pictureless day, too bad.

Moving away from Mayavi2 I have been moving towards ActionScript and Flash. Why this switch? Well Mayavi2 is a real cool scientific visualization tool, but most of the focus is on on-the-fly visualization of data that is dense, but also a computer that is powerful. This does not really serve useful for the purposes I would want to move towards. Another major drawback is that I haven’t been able to find a way to export the visualization I have made in Mayavi2 to another format than a static PNG.

Finding out that the FLARToolkit almost does all the things I am moving towards I swallowed my pride and installed the Flex SDK. This is a first time for me to do scripting in ActionScript, but I must say it has been most rewarding and a lot more useful and powerful than I have always judged Flash to be, always shying away from it, because I hate website that use Flash to do things that CSS and HTML can easily perform.
Diving into ActionScript and Flash 3D engines has been educational, I’ve been able to rotate cows, which was fun, as well as a few other things. The FLARManager allows you to use 3D engines such as PaperVision3D, Sandy, Away3D and some others that don’t surface for now.

So that was good. But then the other aspect. Trying to get a GIS file to 3D. You’ve got to be shittin’ me why should this be so hard.

Trying to get a GIS file (even raster images) into a 3D format seems to prove more difficult than anticipated. It is not so much that I do not know how to visualize or how I would like it to be, but I keep running into these stupid data type issues. I am almost frustrated up to the point of trying out a red brick vs. LCD screen collision.

ArcScene
ArcScene exports 3D to VRML (.wrl) format, which is almost completely useless (unless someone is able to persuade me otherwise, I will stick to this view). ArcScene is however by far the easiest way to go from a flat surface to a nice 3D visualization. This bothers me to some extent, because of my primal urge for open-source software to be better.

Step-by-Step
So what I am going to try now is to export the raster image I have to a PNG file and to load that in Blender. Blender can export to X3D (the successor of VRML) but not import from it (another point of frustration). Trying out all of these different import/export tools almost made me parse the VRML myself and spit out a COLLADA file. Maybe this is still the best way to go. So if I do I will post the script here.

For the time being I am going to load the greyscale into PNG to create a height map in that using the same technique as the last post featured, and export that to COLLADA or another suitable format for pv3d or something. Let you know how it goes (hopefully with nice pictures).

P.S. check out this awesome portfolio: mrdoob.com

Processing images with GDAL and Mayavi2

Ok, this took a while but the results are not bad. So I guess it’s worth it.

First of all, a big thank you at Chris Garrard who is very possibly unaware of the fact that I have been using his classes on Python and Open Source GIS. They give a good introduction into using GDAL with NumPy.

So after struggling a few consecutive days (frustration peaked!), I managed to load a Digital Elevation Model into the Mayavi 3D visualization tool. It’s still a bit rough and the peaks are a bit pointy, thus buildings look a bit weird, but still it’s another little step.

I first tried a Blender Tutorial that helped me out making this image:

Blender output

Then after that I figured that is a bit much work for something that in my opinion can be done more efficiently. The steps are still a bit stupid, because I exported a PNG of the DEM visualized with grayscale and then imported that into gdal and displayed that. So it’s still a bit dumb. And it isn’t really representative of the data types that a DEM contain. But the DEM itself only gave very weird visualizations, possibly because of the detail of the file as well as large negative values for NoData values.

Anyway it is a start:

Mayavi2 Output

Developments so far

Ok. 3 weeks down the road of my minor thesis and what do I have to show for? Well, a lot of paradigms and concepts for starters. Furthermore I have never been able to finish a document like research proposal so quickly and of such quality.

But enough of that. I have more or less abandoned the idea of a mobile application for now. Because having a mobile application like Junaio scores high on the cool list, it doesn’t really in the sphere of performance and handling large data sets. The preference might go to having a standalone or online application that uses a webcam to model the information on screen.

Most interesting technologies so far:

1. The awesome Unity 3D game engine. Apparently possible to combine with SSTT – AR tool
2. The extensive Python libraries of VTK and Mayavi2 (installing in Ubuntu costs only 10 min.) in combination with GDAL
3. (Open)Gamaray – An AR toolkit similar to UnifeyeSDK (Junaio) by Metaio. Possibility to add location and 3d objects. Very cool.

Another harsh realization is that I have grown too fond of Ubuntu Linux and only now have to deal with an issue that other programmers have been dealing with for ages. Building from source. My goodness. In Mac this feels too dangerous as a mismanaged system can have a lot of missing dependencies and building takes quite a while. In Windows this absolute hell. Searching around for pre-compiled binaries is the way to go, as are packages like Pythonxy. But still, this sucks.

What I have been able to do so far (except for spending days writing and compiling) is importing an Aster Satellite image into the Mayavi2 viewer. Which was remarkably painless and cool. The only thing is that I do not really understand yet what it is showing me. I guess the pixel values of one band have become the Z-axis. Still, cool as Sir Dangles.

View from the middle

View from above

Also importing height maps in Unity3D should be not be too painfull, but this proved a bit harder than anticipated.

Different Directions

It seems that my minor thesis will focus more on the user experience of augmented reality. The EIA’s have raster outputs that can be converted to 3D spaces. These spaces can then be given Locations. The Raster is already georeferenced which would make this easier (I think).

The best method of viewing these things would seem Augmented Reality browsers such as Layar or Junaio. So I would have to brush up a bit on what is possible in which browser. Both are iPhone and Android compatible which is convenient.

In a later stadium it might be cool to add 3d models of ships going to the harbour. (wishful thinking).

Random thoughts Augmented Reality

Another consideration is if using Augmented Reality (AR) how to implement it and what method of going about it.

ARToolkit has different spin-offs that look quite interesting. One of my favourites are FLARToolkit and possibly SLARToolkit. The flash toolkit. But then I am not very proficient in Flash and ActionScript. Also I tend to avoid Flash as much as possible as well as writing stuff in C#. But the office I work in makes it easier to use C# then any other binding. Also there is a C# binding of OGRE which is quite well developed. OGRE is an opensource 3d gaming engine Which made me think: HTML5. Is there a solution for that?

Also OSGARToolkit looks interesting. It uses a well developed opensource C++ based 3d graphics toolkit OpenSceneGraph. Which itself has a GIS project going as well. (osggis and osgearth). A quite detailed and beautiful Flight Simulator – FlightGear – is also based on OSG. An interesting project based on the OSG library is Vizard. A library in Python (yey!) to create Virtual Environments. The only major drawback is that it costs a fortune.

For some cool examples of FLARToolkit: rawkes.com.

The most viable of options seems to be the OSGARToolkit. It is free and above all it has quite good thorough documentation. To get started and to continue the work a bit. Just need to learn a bit of C++ (oh lord) and 3D modeling. This might be a long insomniac season.

Considerations on which Gaming Engine to use

A quick update on my project of using a Gaming Engine to process GIS data. After some googling and reading other peoples blogs and workspaces it seems that some engines are more fit than others.

Tim Germanchis et al. (2005) has done a brief overview of different game engines for this particular usage. Amongst which are the Quake III, Unreal Runtime 2, CryEngine and others. The CryEngine apparently suited the writer the most. As it was easy to learn, stable and the most powerful.

However according to the gis-game workspace (coming up in almost all of my searches on Google) this Engine is not very favourable. By the way if anyone knows who these guys are, I would love to come into contact with them. But found it kind of difficult to contact them.
One major objection I personally have with CryTek is that it is a Windows only Software. As I will have to do most of my development behind a Windows I will probably would be in the temptation of cutting my lifeline short. But it does make it easier to user CryEngine as well.

I also saw that CryEngine 3 is coming out. Maybe my Prof would like to help me to get a Educational license to start working on that.

In relation to open-sourcing and stuff: ioquake3. The well-known quake3 engine has been released under the GPL license as ioquake3. So maybe that is worth giving a try. Although it is a bit older. But then again, science seems to lag in this field anyways.

One of my major concerns is the processing of GIS data in 3d. I hope that I can find a way with Python and bash maybe to automate the modeling of the 3d objects. Although that might be wishful thinking.

My other concern I will discuss tomorrow. As it is time for me to do some studying for my courses.

Edit: this looks interesting for DEM and Blender things

Usability and 3D in Geovisualization

For my minor thesis I am thinking of a project for processing Environmental Impact Assessment in a way that is more easily comprehensible. For this project I intend to use Augmented Reality. Hopefully some form of 3d visualization. And I am looking at the possibilities of using gaming engines.

First and foremost, this is a preliminary research. I am not in any way an expert. And only just learning about the amazing possibilities.

Over the last few days I have been looking at different scholarly and not so scholarly papers in search of some good information on the use of game engines in 3d visualizations of GIS. The search has been quite fruitful but not conclusive. Aren’t these searches always.

Anyway. This is more or less a note to self. To remind myself what I’ve found and where. But if you’re interested please follow this blog to see the outcome of this project.

Visualization using Game Engines is something that has over the last decade caught the attention and imagination of more people than just me. Not surprisingly because the gaming industry is leading the field in wonderful visualizations and magnificent displays of Artificial Worlds and the interaction of the user with it.

1. Fritsch and Kada – Visualization Using Game Engines
Dieter Fritsch and Martind Kada of the Institute for Photogrammetry in Germany has written an introductory overview on different game engines and their applicability and capabilities for visualization in 3d. His conclusion clearly states the possibilities but also the complications. These mainly consist of data format. Common GIS data patterns are very different to most 3d data types. Polygons and points are usually a bit more complex more on this in number 3. A project called GISMO apparently is quite capable of rendering large data sets and are quite user friendly.

2. Herwig and Paar – Game Engines: Tools for Landscape Visualization and Planning?
Interesting conclusions and research on the increased involvement of stakeholders and planners. Not necessarily a very technical paper. But good for reference.

3. Scianna and Ammoscato – 3D GIS Data Model Using Open Source Software (2010)
An interesting article on using the Python bindings of Blender to process GIS data. They start of by stating why they chose to use Blender for the research and not other open source 3d software. There is quite a difference in data types as already mentioned between usual 3d rendering and design software and GIS data types. Unlike GIS types points and lines do not form the primitives of 3d software.
Blender and the GIS RDBMS PostGIS are then linked through a Python script. And data is sent back and forth. Python scripts are available online: Link does not seem to work

4. Slocum et al. – Cognitive and Usability Issues in Geovisualization
This is a more academic work on cognitions and user friendliness. Different methods of working with Geovisualization are proposed. Good for reference and Usability things.

5. MacEachran (editor) – Exploring Geovisualization
Huge collection of articles on Geovisualization. From Usability and Cognition, to Data Handling and 3D interfaces. Chapter 28 was quite useful on Usability.

Of course much more literature has been leafed through, but these are some of the articles or books that I found worth mentioning in the introduction of my proposal. Next time some more literature overviews and then hopefully some conclusions.