Point Cloud V4 Issue 2 (Nov 2018)



The Point Cloud, AWARE’s Electronic Newsletter
Vol. 4, Issue 2. Date: 11/13/2018
Previous Issue

Not displaying correctly?

View this email in your browser

Time flies! Halloween is over and we are halfway through our fourth year. There has been a lot of progress in AWARE research recently. Six researchers – nearly a quarter of AWARE’s students and PDFs – are expected to finish up by the end of the year. This is an opportunity to get a leading edge researcher on your team, as some of them may be looking for work. Last of all, congratulations to Tristan Goodbody and Joe Rakofsky, who have both had papers recently accepted for publication.

In this edition:


Research Snapshot

AWARE Videos

AWARE Update



Feature Researcher

Joe Rakofsky earned his Hons. BSc. in environmental sciences from McGill University in his home city of Montréal, QC. In the final two years of this degree, Joe dabbled in labs working as a research assistant and focused on a wide range of topics: from a micrometorology lab that assessed carbon fluxes in wetlands, to a soil biogeochemistry lab where he assessed nitrogen fixation rates in relation to water table depth. Ultimately, he solidified his passion for spatial analysis while working in Dr. Jeffrey Cardille’s remote sensing lab. He worked on a project that used spectral data from Landsat to quantify coloured dissolved organic matter content in lakes of the Abitibi-Témiscamingue region of Québec.

In September, 2016, Joe moved to Vancouver, BC to start as an MSc. Candidate in the Integrated Remote Sensing Studio under the supervision of Dr. Nicholas Coops at the University of British Columbia (UBC). Joining the AWARE project allowed him to shift his focus from peatlands to forests. For his research on Q12 of Theme 2 of AWARE Joe uses airborne laser scanning and image-based point clouds, acquired eight years apart, to model height growth estimates and to assess the impact of plot-level mortality on the predictive accuracy of this model. He then uses the same point clouds to predict height growth across a field site near Slave Lake, AB in order to observe growth patterns in relation to stand-level ecological variables. Since starting at UBC, Joe has also run workshops on forest structure modeling from point cloud data. He conducted these workshops in-person for some of AWARE’s industry members in Newfoundland, New Brunswick, Québec, Alberta, and locally in British Columbia.
While outside of work, Joe enjoys studying Mandarin Chinese, assembling idle origami projects, and running, as well as hiking, camping and other outdoor recreation.


Research Snapshot


TLS OcclusionCorrection using Relative Density Index (RDI) for Forest Structure Estimation

Abdelmounaime Safia is originally from Algeria, where he studied geodesy and remote sensing at the National Center for Space Techniques. There, he obtained his master’s (M.Sc) and his first doctoral degree. He moved to Canada to continue his research where he obtained a second Ph.D. in remote sensing at Sherbrooke University. In 2016, Abdelmounaime joined AWARE to work on the application of Terrestrial LiDAR System (TLS) in forestry (Q19) at the individual tree (Theme 3) and plot levels.

For his research, Abdelmounaime is working on improving methods to correct for occlusion in TLS. Occlusion results from forest material blocking the laser scan, causing the volume behind the blocking material to appear empty when that may not be the case. Occlusion is proportional to stand structural complexity, stand density and distance from the scanner. This distorts the distribution of points and consequently adds inaccuracy to estimates for metrics like 3d vertical distribution. The occlusion error can be reduced by performing multiple scans but doing so introduces oversampling problems.

The relative density index (RDI) was proposed to overcome the above-cited limitations based on a voxel representation of the point. The RDI is a normalization of the number of TLS hits inside each voxel by the number of potential TLS beams that cross that voxel. Its main limitation is related to the assumption of a regular spherical TLS shooting-pattern. Due to scanner imperfections including vibration and rotation sampling irregularity, the laser shooting pattern does not follow a regular spherical distribution. The difference between the effective shooting pattern and a theoretical misestimates 3D forest component distribution.

Abdelmounaime proposed a new strategy for estimating the 3D vertical distribution of forest material based on a modified version of the relative density index (RDI). For that, the modified RDI combines both filtered and unfiltered TLS data: instead of using only TLS hits that correspond to exact locations of forest materials, he takes advantage of sky points and mixed points, which are usually considered as noise. The idea behind using the three types of point cloud is to deliver an exact 3D distribution of forest material based on the TLS shooting architecture. Results shown in the figure below quantifies the gain when adopting the modified RDI. Recently, Abdelmounaime contributed to the effort of converting the RDIS into plant area density (PAD) which is taught to be more related to forest biomass. Abdelmounaime also introduced some crown metrics based on voxel representation including crown surface exposed to light, crown gaps and density based on the RDI and the PAD indices. Abdelmounaime and co-authors are working on a manuscript based on these results. He and his co-authors are also using the RDI and the PAD to map forest understory in boreal and Mediterranean forests.


AWARE Videos

Video summaries of AWARE research are now online. The videos are short summaries of the research question and the researcher’s thoughts about being part of AWARE. We have a few of the research questions up and will add more as more research questions finish. Check it out here!



AWARE Update

In August, a portion of our 4th year budget was held back due to less in-kind support than anticipated and, as a result, the plan for two more AGMs has been modified and we will have a larger final AGM at the project end in early 2020. Increasing in-kind support is critical for prevention of further holdbacks. While we are not out of the woods yet, we continue to work with our industrial partners to increase support and mitigate the potential for another holdback next year.

Copyright © 2016 University of British Columbia, All rights reserved.