The future of flightsim scenery?

[toprob]

This was linked from the Avsim forum today. This Swedish site includes a Google-type photomap with a twist -- it includes a high resolution mesh which defines the buildings and trees as well. I've always thought that this would be the way to go with flightsim, removing the need to model city buildings, you just need an appropriate mesh. Not quite possible with FS's limited mesh resolution at present, but shows a lot of promise.
Here's a shot from Stockholm -- although a still image, on the site you can rotate it around and scroll all over the city.

[Ian Warren]

WOW .. Thats definitely helicopter scenery , guess be a few years even before were going to see that , or is it

[creator2003]

Yeah wow that looks pretty good from that distant but having a helipad harden on it would feel alittle strange ,but all the same wow pretty cool ...
thanks for the link rob

[Charl]

Bit like 3D printing I suppose, wonder how they texture the vertical faces?

[spongebob206]

Can't wait for the technology to progress this far. Hopefully only a few years to wait.

[ardypilot]

Absolutely incredible- I wonder how the file size of this sort of scenery would vary compared to traditional BGL scenery files?

[Timmo]

The Tarawera scenery I released used the same sort of technology: Lidar

When we get the datasets, they usually come in three parts (all from the same source data)- All laser strike points, ground points (all laser strike points except trees, buildings etc) and non ground points (laser strikes that fell on trees, buildings etc).

When we make the terrain from this we usually get dont include the non ground points (since, for the most part the Regional council is interested in the shape of the land and not so much the buildings). We can/do use all the points for such thing as vegetation change detection, building change detection etc.

So....the data is there now.

Charl raises a good point: How to make the building side texture as an aerial photo doesnt 'see' the sides of a building much....Im not sure how they did it here but there are a number of options: aim/fly the Lidar unit at quite a low oblique angle and capture photography at the same time or use a ground based lidar survey. Some of the ground based units allow capture of RGB values for each laser strike so they can 'paint' a 3D picture.

Personally I dont think 'full mesh' is the way/will be the way to go as the models arent optimised- There is, however, software that will convert messy point clouds into bonafide 3D models with proper faces. I think this will be the way with councils running ground base LiDaR surveys to semi automatically re-create their cities.....its still expensive though.

[chickenman]

QUOTE (Timmo @ Jun 3 2008, 07:37 PM) <{POST_SNAPBACK}>

The Tarawera scenery I released used the same sort of technology: Lidar

When we get the datasets, they usually come in three parts (all from the same source data)- All laser strike points, ground points (all laser strike points except trees, buildings etc) and non ground points (laser strikes that fell on trees, buildings etc).

When we make the terrain from this we usually get dont include the non ground points (since, for the most part the Regional council is interested in the shape of the land and not so much the buildings). We can/do use all the points for such thing as vegetation change detection, building change detection etc.

So....the data is there now.

Charl raises a good point: How to make the building side texture as an aerial photo doesnt 'see' the sides of a building much....Im not sure how they did it here but there are a number of options: aim/fly the Lidar unit at quite a low oblique angle and capture photography at the same time or use a ground based lidar survey. Some of the ground based units allow capture of RGB values for each laser strike so they can 'paint' a 3D picture.

Personally I dont think 'full mesh' is the way/will be the way to go as the models arent optimised- There is, however, software that will convert messy point clouds into bonafide 3D models with proper faces. I think this will be the way with councils running ground base LiDaR surveys to semi automatically re-create their cities.....its still expensive though.

Timmo has access to cool toys! We are starting to look at LIDAR for a number of jobs as well, with a view towards using it for 3D of buidings and trees in an urban environment (regional council should look at roof catchments for runoff at least, Timmo). The issue we have is negative detail such as eaves and the like but the effect is generally the same. The other issue is where there are trees or other obstructions on faces of buildings such as ivy growing up walls. In general LIDAR is combined with digital imagery, it's the only way you can proof such large areas without having to walk around the area of interest.

The main labour in using lidar and terrestrial laser scanning for buildings is getting the faces all cleaned up and getting rid of all of the point data. We use a Leica HDS3600 with Cyclone Model software for getting cad objects which is pretty good but best for complex surveys like oil refineries which have a lot of distinct seperate objects. It doesn't work so well for flat sided buildings. The company doing the lidar or scanning will generally do the modelling for you. Did a cylinder water tank with laser scanning and have found a few issues that would raise their heads in scanning aircraft for 3d models.

LIDAR definitely rocks for terrain BUT you get way too much detail. I had a quick look at using LIDAR data for 3D terrain modelling with aerial photos but the PC fair screamed at the effort. I think that LIDAR is definitely the way forward for FS scenerry design but you don't have the ease of access to data. And at the moment it is sooooooo expensive compared to current methodology such as photogrammetry.

Jamie

[toprob]

QUOTE

When we get the datasets, they usually come in three parts (all from the same source data)- All laser strike points, ground points (all laser strike points except trees, buildings etc) and non ground points (laser strikes that fell on trees, buildings etc).[/quote]

My vague idea for FSXX was to have access to two datasets, and be able to flag city areas where highres photo imagery exists. You wouldn't want your airport buildings defined by mesh, and you certainly wouldn't want bumps where no photo exists, so this would just be an extra feature similar to today's local photoreal scenery. You might need to build actual models for major landmarks and landable surfaces, but the idea of modelling a whole photo-real city, with much greater accuracy, automatically and without autogen has a lot of appeal.

Ok, a lot would have to happen in FS for this to work -- the sides would need a separate image -- but since much of the data already exists today, it would be a sensible direction to go.