Stem Classification and Modelling from Lidar for a Semi-Automated Forest Inventory

Abstract

Classifying and modelling tree stem characteristics such as tree height and diameter is a major challenge in remote sensing. Their accurate modelling using lidar technology would be one step forward to automation of forest inventory. Forest inventory is the process and product of a survey that assesses a forest resource, and is of paramount importance in forest management planning. Forest resource refers to the renewable or non-renewable material required or needed, such as timber products that can be drawn on, or anything for which there is a perceived present or future use. While airborne lidar technology offers the advantage of quick and efficient large-area forest coverage, the reliability and precision of the measurements depend on the validity of assumptions made in models applied to derive inventory-relevant parameters from point clouds. The aims and objectives of this study are to address the gaps in research areas of tree stem extraction, taper diameter estimations from species-specific trigonometric variable-form taper equations, and more robust diameter at breast height estimations using circle fitting algorithms, in order to develop a semi-automated stem detection and extraction technique from lidar point cloud data. Semi-automated stem detection and extraction techniques utilising terrestrial lidar point cloud datasets have been tested and validated against manual field measurements including destructive sampling of trees, and satisfactory results have been obtained for trees that are sufficiently scanned from multiple set-ups for the algorithms to detect the diameter. Four circle fitting algorithms: Taubin, Pratt, Polar, and Polynomial, are tested for estimating diameter at breast height measurements from thinned data acquired from the terrestrial lidar scanner Echidna™. An automated method to detect and separate individual trees and calculate tree positions and tree heights, as well as evaluate the quality and accuracy of the measurements based on terrestrial lidar data are also presented in this study.