How it works?
The scanning time suitable to produce dense pointclouds is 3 hours for 10 hectares. Large areas (>100 ha) can be covered by successive, overlapping scanning.
The AI algorithm begins at a large nucleus of points with high density and then grows by accretion until it meets neighboring trees.
VirtSilv calculates for each individualized tree the x, y, z position, the volume determined based on a unique allometric equation, and the group of species (coniferous/deciduous).
The raw data generated during the scanning process enables the visual identification of individual tree structures, but they are not yet quantitatively differentiated. To create individual raw material for digital twin, VirtSilv first separates the ground from the trees, and then it reconstructs each tree separately.
The algorithm takes three steps to estimate each tree’s footprint simultaneously. The algorithm begins at a large nucleus of points with high density and then grows by accretion until it meets neighboring trees. As a result, the operator is given many options to customize the algorithm and is given the option to change data sets according to their needs. The average processing time of segmentation was 30 min for a 1-ha plot.
Digital Twinning Process
When all of the individual tree segments are identified, the remaining task is to recognize tree trunks and model their numerical dimensions on a simple and flexible basis, thereby giving the potential for the digital twinning process. To overcome the limitations of current techniques, VirtSilv algorithms are designed around the following principles:
The trunk shape of segments of sufficiently small height can be approximated very well by inclined cone trunks;
The vertical projection of the data obtained from segments of sufficiently small height can be approximated by a ring of points with relatively high density;
Generally, the successive segments in the vertical array are very well aligned, in the sense that the angle and bending of each segment, concerning that vertical changes are low.