CAD

“What’s the promise of intelligent 3D models?”

I am going to answer this in a different way this time around. Included are a lot of images I created or research I worked on and for which images were generated. My thesis was on visualisation and 3D, so this topic piques my neurons. The promise of 3D intelligent models is very high and likely to happen within the next three years in useable forms. This depends of course upon what you mean by 'intelligent' models.

 

Visualising in 3D has to do with perspective. Where you stand influences what you can see. But, you don't have to see everything in 3D to know something lies within it (see below).

 

 

For a model to be intelligent, means that the model must link to data. The data supports the model not only in terms of geometry, but in terms of attributes and how the model will be described. The greater amount of information we have about a model, then the more intelligent the model is. This is why a big challenge and goal has always been to link GIS / CAD to visualisation. The first two are more closely connected to the database, while visualisation is more closely in resemblance to the camera. Simply put, connecting the two means real data connects to real models and scenery / objects.

 

My earliest ventures into this began through studying forests. I wanted to be able to connect a forest to a database through remotely sensed images, but not the regular way. Instead, I wanted to assess forest stands vertically, from the side. The process went like this.

 

Using a GPS camera (I had a Kodak GPS enabled camera back in 1999) I could generate a series of images from a transect perpendicular to a forest stand. These would allow me to generate stereoscopic pairs. (see below)

Reasoning on the same principles as aerial photography, I was able to generate a series of points (call it a vertical DTM if you like) and then import them into a GIS database. I would triangulate the points (TIN) into a workable continuous surface. At the time I was using ERDAS Imagine for my analysis and could delineate forest structure and assign distances to them from my GPS location with Arcview 3.x for TIN (see below).

Eventually one ends up with a forest that, for all intents and purposes, has timber diameters and heights that are measureable, and spatially represented. The logic here was to re-visit the GPS locations over time and perform the same analysis. In effect, these would become permanent sample plots. Through creating several site visitations of data, the whole lot could be analysed in a GIS over time. This would allow timber growth, forest structure and other assessments to be made.

 

As this work proceeded, the idea of having more intelligence into a 3D model became more interesting and challenging. This led to ideas like coupling soil sample grids to landscapes (see below).

And later this would be presented with actual tree species libraries on the landscape over the soil analysis areas relative to the terrain. For this, the 3D Nature company was well ahead of the pack in terms of coupling databases to real visualizations. We were creating visualisations 8 years ago and dynamically linking the models (trees, houses) to the landscape and through one product (NatureView, again 3D Nature) we could fly through them. This was an important step for understanding the value of time-lines, which, today, both ESRI and Google include in their Virtual Globe products. For me, it led to the idea of creating landscapes where forest fires were happening (I used to fight forest fires). If the fires could be sensed then presented on a terrain, then where to locate resources and drop water would be more easily understood (see below).

 

Intelligent models need to connect visual data and information to a realistic setting. In developing a 3D model for the Tiergarten Park here in central Berlin, I tried to place images and maps into a DTM. The idea was to be able to walk through the landscape with intelligent information being presented as you go (see below).

Later we attempted to make an 'emotional landscape' based on one person presenting an idea to another (presenter-presentee). The tough part here was being able to assign an accurate (useful) value to the emotions and responses – we failed (but see example below). This explains why any input for people needs to involve sociologists into the model making and other folks in those domains.

 

 

 

Coming full circle, I simply wanted to present a realistic scene generated from a databases that I had control over and analyse. The following picture is totally computer generated but backed up to a database for terrain, vegetation and water features (see below).

 

 

 

To me, an intelligent model is one that is highly and closely connected to the database. It can be in the past, present or future, as in a simulation. But it must connect to a database. There are some important things to understand about intelligent models.

 

• The database in use must be able to scale and perform literally thousands of computations per second if real-time intelligent models are to be realised. For this reason, I see limited use of intelligent models for moble applications because of the huge amounts of data needed for real-time operation. A more workable solution is to have regional centers of intelligent computing with super computers that feed to a smaller and wider grid in more focused apps.

 

• GPS are critical to intelligent models. They locate all things on the landscape and enable high resolution models to be more precisely placed. I am not thinking only of environmental applications here, but also utility and infrastructure applications where real and planned features can be visualised.

 

• Sensors will play an increasing role in 3D intelligent models. This is one of our focus at V1 Magazine for this reason. Sensors will enable automation of data and intermediate processing of information locally before integration.

• Today, image compression is an important issue if you are considering 3D intelligent models. Images are raw forms of data and they also provide photo-real presentation for draping. Far and away, their value for data is more important, especially if they are sensing the same place with higher visitation rates and resolutions.

 

I don't see military visualisation and 3D intelligent models as any different than other types of visualisation. I think there is more data to process for military applications, but the challenges lie mostly in packaging or 'bundling intelligence' for use. Without going much further here, military 3D model intelligence offers the greatest prospect for coupling sensors, visualisation, GIS/CAD and 3D models together in unique and new ways for use. The concept of 'personal military geo-space simulation models' would be interesting to explore also, especially in an inter-connected way.

The promise of intelligent 3D models is better decisions, saving money and higher levels of understanding.

Finally, I can't see how more and better 3D intelligent models can be developed without high levels of data quality. Average data and analysis will only provide mediocre, non-useful and non-functional decisions to be made.

 

Chin up and keep thinking on it...

 

 

Read what Matt Ball has to say on this topic here.

 

Related Posts

1. “What is the promise of the model-based future for multi-disciplinary planning projects? ”
2. Photosynth as a Platform for Digital City Models
3. “How will technology evolve so that city models become the new base map?”
4. 3D Service Oriented Interoperable City Models
5. “Will digital city models become one über model or separate models for specific domains? ”
6. Subsurface 3D Models for Mining
7. Bentley Releases V8i for the Intelligent Infrastructure Cycle
8. 3D Rural Community Nutritional Models
9. Denmark Models Integrates and Shares Spatial Data Across Government
10. Digital Terrain Models: GPS or Aerial-Satellite Sensing?

© 2007 Vector One.