Multi-3D-Window Dead Tree Detection Of Dead Standing Eucalyptus Camaldulensis From Voxelised Full-Waveform Lidar Data For Tackling High Differences In Native Forests
Morressier, 2017
Online
unknown
Detection of dead trees is an important for managing biodiversity in native Australian forests. Most of the previous work on dead standing trees detection performs single tree crown delineation before health assessment. Nevertheless, classifications at tree level, while working with native forest is a challenge for multiple reasons: big spatial resolution, variance density of trees, different tree heights and sizes. Tree crown delineation is usually done by detecting local maxima from the canopy height model (CHM) and then segmenting trees using the watershed algorithm, but Eucalypt trees has multiple trunk splits making tree delineation difficult. Shendryk et al, 2016, published an interesting Eucalyptus delineation algorithm that performs segmentation from bottom to top, but pulse density was 36 points/m2 around forested areas (expensive to acquire for big spatial resolution). Miltiadou et al, 2018, attempted detection of dead Eucalypt trees without tree delineation using 3D-windows and showed that it is possible, but the methodology can be improved. The presented work, takes that research a step forward and uses multiple 3D windows charactering dead trees. A random forest classifier, a weighted-distance KNN algorithm are used to create a 2D probabilistic field for each 3D-window size. Then the results are merged, and the locations of the dead trees are predicted. It is shown that the multi-3D-window approach improved the results of the original research published work in 2018.
Titel: |
Multi-3D-Window Dead Tree Detection Of Dead Standing Eucalyptus Camaldulensis From Voxelised Full-Waveform Lidar Data For Tackling High Differences In Native Forests
|
---|---|
Autor/in / Beteiligte Person: | Miltiadou, Milto |
Link: | |
Veröffentlichung: | Morressier, 2017 |
Medientyp: | unknown |
Sonstiges: |
|