Improving TLS-based stem volume estimates by field measurements

Timo P. Pitkänen, Pasi Raumonen, Xinlian Liang, Matti Lehtomäki, Annika Kangas

Research output: Contribution to journalArticleScientificpeer-review

9 Downloads (Pure)

Abstract

The prediction of tree stem volumes has conventionally been based on simple field measurements and applicable allometric functions, but terrestrial laser scanning (TLS) has enabled new opportunities for extracting stem volumes of single trees. TLS-based tree dimensions are commonly estimated by automatized cylinder- or circle-based fitting approaches which, given that the stem cross-sections are relatively round and the whole stem is sufficiently covered by TLS points, enable an accurate prediction of the stem volume. The results are, however, often deteriorated by co-registration errors and occlusions, i.e., incompletely visible parts of the stem, which easily lead to poorly fitted features and problems in locating the actual treetop. As these defects are difficult to be controlled or totally avoided when collecting data at a plot level, taking advantage of additional field measurements is proposed to improve the fitting process and mitigate gross errors in the prediction of stem volumes. In this paper, this is demonstrated by modelling the stems first as cylinders by only using TLS data, after which the results are refined with the assistance of field data. The applied data consists of various field-measured stem dimensions which are used to define the acceptable diameter estimation limits and set the correct vertical extents for the analyzed tree. This approach is tested using two data sets, differing in the scanning setup, location, and the measured field variables. Adding field data improves the results and, at best, enables almost unbiased volumetric predictions with an RMSE of less than 5%. According to these results, combining TLS point clouds and simple field measurements has the potential to produce stem volume information at a considerably higher accuracy than TLS data alone.

Original languageEnglish
Article number105882
Number of pages10
JournalCOMPUTERS AND ELECTRONICS IN AGRICULTURE
Volume180
Early online date2020
DOIs
Publication statusPublished - Jan 2021
Publication typeA1 Journal article-refereed

Publication forum classification

  • Publication forum level 2

ASJC Scopus subject areas

  • Forestry
  • Agronomy and Crop Science
  • Computer Science Applications
  • Horticulture

Fingerprint

Dive into the research topics of 'Improving TLS-based stem volume estimates by field measurements'. Together they form a unique fingerprint.

Cite this