Automatic Generation of 3D Building Models from Point Clouds - Geoinformatics for Intelligent Transportation - page 116

5 Conclusions

The software has a user-friendly graphical interface that is quite intuitive and allows
users to view analyzed data in several options. Data processing is very well
described in the Help section. The main advantage of this software is the possibility
of automatic data processing, which is controlled by user-de ned parameters.

However, the description of the parameters is not very detailed in the Help of the
software. Options are described in general and very tersely in the text. A user does
not have suf

cient insight into the internal functionality of the software.
The processes of the point cloud classi cation and modelling of buildings are

very robust but they are not too sensitive to changes of optional parameters. This
can be considered as an advantage and disadvantage at the same time. The modi-

final result.
For this reason, recommended settings were used in most cases. An implemented
method analyzes the data locally and without the use of any knowledge base of
buildings or roof shapes. The building model is created by a group of smaller parts
(irregular polyhedrons). The software does not create real building models, but only
groups of spatial primitives.
The main weakness of the software is its necessity to use high-quality input data.
Processing of different data sets showed that for achieving high-quality building
models, it is necessary to use a very dense point cloud as an input (20 points per
square meter and more). Unfortunately, point clouds with this density are not
typical. The missing knowledge base of building and roof shapes, and a poor
control of the implemented algorithm make the processing of sparse point clouds
dif

fication of some parameters did not lead to the desired changes in the

cult. Applying a sparse point cloud (up to 10 points per square meter) results in
the creation of building models with a lower quality.
At the end of this paper, it should be mentioned that a detailed 3D city model
consists of an accurate digital terrain model and high-quality 3D models of
buildings and grown vegetation and all objects alongside roads. The data will
become a normal part of the visualization of such space to allow designers to
enlarge their data sources and geographic data for further processing and modelling.
They will enable specialists to control and even improve conditions of the road
traf c to be safer and passable. All these spatial data have a great importance for

intelligent transportation, therefore they are quite important.

References

1. Haala N, Brenner C (1997) Generation of 3D city models from airborne laser scanning data.
In: Proceedings 3rd EARSEL workshop on LIDAR remote sensing on land and sea, Tallinn,
Estonia, 17 - 19 July, pp 105 - 112
2. Haala N, Anders KH (1997) Acquisition of 3D urban models by analysis of aerial images,
digital surface models and existing 2D building information. Integrating photogrammetric
techniques with scene analysis and machine vision III, Orlando, FL, United States, 21 - 23