Automatic Generation of 3D Building Models from Point Clouds - Geoinformatics for Intelligent Transportation - page 107

Automatic Generation of 3D Building
Models from Point Clouds

Vojt ch Hron and Lena Halounov

ě á

Abstract Point cloud is a product of laser scanning (terrestrial/airborne) or it can
be derived from automatic image matching. Both techniques are very modern and
progressive methods of non-selective collection of spatial data. The representation
of buildings through point cloud is not appropriate for many applications. Handling
with a set of data points, covering large areas is also very hardware consuming. For
these reasons, it is suitable to represent individual buildings as spatial objects, called
3D models. This paper is a review of fully automatic generation of 3D building
models from point clouds. It compares the solutions of various academic institutes
and analyzes current commercial software products that process this task. In this
work, data point clouds collected by airborne laser scanning will be used as an
input. A major in uence on the generation of 3D building models have the density

and quality of the point cloud, which are determined by scanning parameters. For
this reason, various input datasets will be tested.

fl

Keywords 3D building models Point clouds Airborne laser scanning LiDAR

1 Introduction

Spatial data are very popular and in demand nowadays. Due to its great popularity,
this data type and high requirements for the accuracy and topicality lead to their
more frequent acquisition. The acquired data have a great information potential.
Manual interpretation of spatial information is in fact extremely time consuming
and it is impossible to repeat it with the same result due to the human factor. Fully