Speaker: Matthew Wozniak

​Bio: Matt is a PhD student in the department of Geology and Geography at West Virginia University. Prior to this he obtained a master’s degree in computational biology from Rutgers University. His interests are in developing methods for improving the utility of terrestrial LiDAR scanners for forestry applications, specifically when it comes to mapping fuels and ecological traits related to fire-spread.

Webinar: The ability to segment out individual trees and shrubs from plot level scans acquired through LiDAR collected data is critical for advancing the utility of the technology, and an assortment of algorithms have been developed to do so. While many do a decent job of identifying individual trees in well-spaced stands, they fail to properly segment trees and shrubs in dense stands with complex structure while preserving the necessary data. We are working on an algorithm and associated workflow which does not automatically segment everything, but utilizes a graph theoretic approach to create workable units which the user can easily identify and merge into individual trees and shrubs using open-source software, preserving all of the necessary data. If we can properly segment high resolution individual components of scans, they can be used to train machine learning algorithms to identify tree species identity and other
useful trait information for parameterizing models of fire spread. Along with this, we are also developing a method for estimating canopy cover and canopy closure using terrestrial LiDAR scans, also only using open-source software. I will discuss our progress with these projects thus far and what we hope to accomplish.