Pointclouds of forests acquired with terrestrial LiDAR (TLS) contain information on structural properties. Currently, a few segmentation algorithms are able to identify and isolate individual trees from the understory.
In temperate forests, the tree segmentation occurs with little or no further post-processing (as in “cleaning the tree from understory”). However, in complex structured forests, such as in tropical forests; post-processing it is believed to be a must-have step before measuring tree parameters (such as DBH, tree height, crown diameter), and estimating volume. This step is done purely manually and is time-intense.
Segmentation algorithms are proved to assist in the location, identification and extraction of trees in tropical forests, until some extent.
Still, some tree pointclouds are needed to be “cleaned” before estimating volume. Cleaning means removing points that do not belong to the tree. Much on the cases are branches of other trees, lianas, epiphytes plants, etc. We believe that by, not removing those points, we are overestimating tree parameters and the volume of the tree.
The goal of this project is to evaluate that if intensive/average/mild tree cleaning affects the estimation of tree parameters and tree volume. We provide the student with a segmented tropical forest plot, in which the student will choose trees which needed post-processing.
Understanding the variability of the tree parameters and volume is an indicator of the need and the intensity of the “cleaning” for tropical trees. This understanding plays a relevant role when scaling from individual trees to plot level indicators.
- Review literature on state-of-the-art tree segmentation algorithms
- Identifying segmented trees which need to be post-processed
- Create a cleaning framework to understand the severity of cleaning
- Estimate and evaluate volume based on different type of tree post-processing.
- Raumonen, P., Casella, E., Calders, K., Murphy, S., Åkerblom, M., Kaasalainen, M., … Kaasalainen, M. (2015). Massive-Scale Tree Modelling From TLS Data. ISPRS Annals of Photogrammetry, Remote Sensing and Spatial Information Sciences, II-3/W4(March), 189–196.
- Scripting skills (e.g. R, Python, MatLab, LaTeX) are a preference
- Basic knowledge of Cloudcompare software Completion
- GRS-32306 Advanced Earth Observation
Theme(s): Sensing & measuring