A voxel-based technique to estimate the volume of trees from terrestrial laser scanner data
Other literature type
(issn: 2194-9034, eissn: 2194-9034)
The precise determination of the volume of standing trees is very important for ecological and economical considerations in forestry.
If terrestrial laser scanner data are available, a simple approach for volume determination is given by allocating points into a voxel
structure and subsequently counting the filled voxels. Generally, this method will overestimate the volume. The paper presents an
improved algorithm to estimate the wood volume of trees using a voxel-based method which will correct for the overestimation.
After voxel space transformation, each voxel which contains points is reduced to the volume of its surrounding bounding box. In a
next step, occluded (inner stem) voxels are identified by a neighbourhood analysis sweeping in the <i>X</i> and <i>Y</i> direction of each filled
voxel. Finally, the wood volume of the tree is composed by the sum of the bounding box volumes of the outer voxels and the volume
of all occluded inner voxels. Scan data sets from several young Norway maple trees (<i>Acer platanoides</i>) were used to analyse the
algorithm. Therefore, the scanned trees as well as their representing point clouds were separated in different components (stem,
branches) to make a meaningful comparison. Two reference measurements were performed for validation: A direct wood volume
measurement by placing the tree components into a water tank, and a frustum calculation of small trunk segments by measuring the
radii along the trunk. Overall, the results show slightly underestimated volumes (–0.3% for a probe of 13 trees) with a RMSE of
11.6% for the individual tree volume calculated with the new approach.