PhD Defence Ms Anahita Khosravipour

Department of Natural Resources

portrait picture_

Title of defence

Detecting individual trees from LiDar


Individual tree detection (ITD) using airborne LiDAR data plays an increasingly significant role in the efficient and accurate monitoring and managing of forest activities. A first step in detecting individual trees using LiDAR data is the generation of a Digital Surface Model (DSM) or a normalized Digital Surface Model (nDSM/CHM) that describes the geometry of the uppermost layer of the canopy. A DSM or a nDSM/CHM is typically calculated by interpolating first-return LiDAR points. Various ITD approaches then identify local maxima in the resulting elevation or height rasters. ITD approaches trying to identify all “true” treetops  are strongly affected by the quality of the LiDAR-derived rasters, which in turn are determined by factors such as the quality of the acquired LiDAR point clouds, the pre-processing, the post-processing, as well as the forest conditions and the complexity of the terrain.

The aim of this thesis is to develop a new approach for generating a high quality LiDAR-derived DSM that improves the accuracy of individual tree detection across multiple forest types and LiDAR point densities. The research in this thesis firstly presents a new “pit-free” algorithm able to create a pit-free CHM raster and efficiently remove those canopy height variations (called pits in a raster and spikes in a TIN) that cause difficulty in detecting individual trees. The algorithm operates robustly on high- and low-density LiDAR data and significantly improves the accuracy of tree detection in comparison to the accuracies achieved using a smoothed first-return CHM. As complex forest terrain presents a challenging problem for the performance of the height normalization step by distorting the normalized DSM (nDSM/CHM), the thesis subsequently aims to quantify the effect of slope on the accuracy of treetop detection in a pit-free LiDAR-derived CHM. To avoid the height normalization step, the research moves on to develop a novel “spike-free” algorithm that can directly generate a DSM (without the need to normalize) at the highest possible resolution using all relevant LiDAR returns. This algorithm considers all LiDAR returns (not just the first returns), while systematically preventing the formation of spikes during the TIN construction process. This spike-free algorithm significantly improves the accuracy of tree detection across multiple forest sites (a temperate plantation in France, a temperate mixed deciduous-coniferous forest in Germany and a tropical rain-forest in Australia) and across different LiDAR point densities. The algorithm offers the possibility of improving accuracy of crown delineation, height estimation, and other biophysical parameters at both a regional and global scale.

Khosravipour, A., Skidmore, A.K. (promoter) and Hussin, Y.A. (co-promoter)  (2017) Detecting individual trees from LiDAR. Enschede, University of Twente Faculty of Geo-Information and Earth Observation (ITC), 2017. ITC Dissertation 309, ISBN: 978-90-365-4396-5.

Full text


Event starts: Friday 29 September 2017 at 12:30
Venue: UT, Waaier 4
City where event takes place: Enschede

previous page
more events

  1. Home »
  2. Organisation »