Campus Model Building Exercise

Summary

Students were requested to model a building from the university campus (twice) utilising both Photomodeler and Google SketchUp. The resulting models would be converted to an OpenFlight format and displayed in the SPLINT visualisation theatre. The objective of this process would be to highlight issues of the relative merits and disadvantages of the modelling packages used, and to demonstrate the conversion processes required to get the models into the Blueberry environment, and issues relating to this process.

Conversion to OpenFlight

Prior to the submission of models by the students, SPLINT investigated processes by which models could be imported to Vega Prime. These investigations using simple example models from each source program identified the routes for model conversion summarised in the following figure.

Results from first session

The models produced by the students were complex in mesh structure, in the main fully textured, and postioned (as requested in this instance) on a georeferenced textured tin surface. Some of the models used externally sourced files to represent key vegetation and generalised structural features such as fire escapes.

The main highlighted issues during lab session:

• Using Google SketchUp method it can be difficult to judge vertical distances and heights on models. One set of models showed this clearly, as the photogrammetric photomodeller method produces accurate building heights.(below).

• Planning and data acquisition – The Lakeside Pavillion had been extended since the overhead photography in Google Earth was taken. Students used oblique imagery as a base for their SketchUp modelling giving an obviously skewed model (below).

• Some sets of models showed the comparative ease of creating realistic looking features with SketchUp models that are more complicated to reproduce in Photomodeller. See the round pillar and chimney features below.

• A white model and a textured model are exported together by SketchUpby default. This was demonstrated by pointing out z-buffering effects on parts of the models below, where there were two identical – but differently textured faces still present.

• There was a difficulty in maintaining model integrity through stages of format conversion. This was demonstrated by the loss of/corruption of textures on some of the imports, and the loss of/corruption of mesh data in some of the more complex models (see below).

• Some of the students used (as requested in the brief) externally created objects in their models, such as vegetation and in the example below, a fire escape. If these objects are intricate in detail they can dramatically increase the complexity of the model. In the case below the complexity was increased to the extent that the exported model could not be loaded into 3D Studio Max.

In conclusion, a realtime VR demonstration of the results of student modelling exercises was used successfully to highlight various important issues of visualisation, building modelling and model conversions.

Contact

If you would like more information about this project please email
andy.burton@nottingham.ac.uk. If you would like more information about SPLINT developments at Nottingham in the areas of mobile computing and 3D visualisation then please email gary.priestnall@nottingham.ac.uk.

[back]