![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
Approximation of volume and branch size distribution of trees from laser scanner data |
VerfasserIn |
Pasi Raumonen, Mikko Kaasalainen, Sanna Kaasalainen, Harri Kaartinen |
Konferenz |
EGU General Assembly 2011
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 13 (2011) |
Datensatznummer |
250050576
|
|
|
|
Zusammenfassung |
Laser scanners can be used to produce point clouds from the object’s surface and thus
to produce a 3D-representation of the object. The point clouds are finite sets of
points of the Euclidean space -3 and they form a sample of the object’s surface. We
analyze laser scanner produced point clouds from trees and propose a method to
automatically extract useful information from the point cloud sample. Particularly, we try to
approximate the total over-the-ground volume and branch size distribution of the
tree.
The method we propose has three separate parts. The first part of the method is to
approximate the mathematical structures of the surface embedded in -3 from the
point cloud sample. The true surface we try to estimate is the result of a stochastic
process and as such is not smooth. However, we assume that it can be reasonably
approximated by a smooth surface. Then the metric, topology, and tangent spaces of
the surface are approximated using the Euclidean metric of the embedding space
-3.
In the second part we identify trunk and branch points from the point cloud
sample. The identification is based on strong assumptions about the data and above all
geometric characterization of the point’s neighbourhoods. Particularly, we define small
neighbourhoods of some radii for every point and then calculate the scatter matrices of the
neighbourhoods. The eigenvalues and eigenvectors of those matrices give geometric
information about the neighbourhoods that can be used to identify the trunk and branch
points.
The third part of the solution is the approximation of the size of the trunk and branches.
The approximation is based on the assumption that the tree can be reasonably approximated
locally as a cylinder. In other words, we assume that the tree can be subdivided into
appropriate subsets which can be approximated well with cylinders of different
radius, length, and orientation. We use geometric information of neighbourhoods to
recognise smaller branches of a bigger branch and ultimately to define the appropriate
subsets. Then we use the total least squares method for fitting cylinders to these
subsets.
We present the problem and the basic ideas of the three part solution. Also challenges
such as incomplete data due to limited scanner positions are discussed. Finally we
present some calculations from real data and compare them to measured result. |
|
|
|
|
|