Discovetree - An Automated Tool To Generate Stem Maps From Terrestrial Laser Scanner Point Clouds

Terrestrial laser scanning (TLS) is increasingly used in forestry to quickly and nondestructively capture a variety of tree attributes such as diameter, height, and volume. However, in order for these attributes to be measured, the individual trees must first be segmented from the point cloud. Tools for manual or semi-automatic tree segmentation are widely available, but a fully automated and generalizable tool does not yet exist. The first step in tree segmentation is the creation of a stem map consisting of the position and size of all trees in the point cloud. We developed a novel stem mapper, "Discovetree" that uses Hough transforms combined with a machine learning algorithm calibrated with field data. Our algorithm outperformed a similar existing tool, TreeLS, at both the tree and stand level. Our study examines the consequences of tree shape on common representations (i.e. circles). The results suggest that stem mappers benefit from being tuned on real-world data and that and that analytical approaches that can represent trees by shapes other than circles may be needed to achieve the same levels of accuracy that in-field or manual mensuration can achieve.

Level thesis:degree_level

Master's Degree

Year dc:date.issued

2021

Author dc:creator

• Hartsook, Theodore Elliott

Advisor dc:contributor.advisor

• Greenberg, Jonathan

Committee members dc:contributor.committeemember

• Weisberg, Peter
• Tavakkoli, Alireza

• hough transform
• lidar
• stem mapping
• tls

Statement dc:rights

• Creative Commons Attribution-ShareAlike 4.0 United States

Handle dc:identifier.uri

http://hdl.handle.net/11714/7969

OAI identifier oai:identifier

oai:scholarwolf.unr.edu:11714/7969