Skip to main content
SLU:s publikationsdatabas (SLUpub)
Konferensartikel, 2003

Detection, measurements, and species classification of individual trees for forest inventory and visualization

Åsa, Persson; Ulf, Söderman; Linköping, FOI; Linköping, FOI; Holmgren, Johan

Sammanfattning

High-resolution airborne laser scanner data offer the possibility to detect, measure, and classify individual trees. Together with aerial imagery this data can also be used for 3D visualization of forest landscapes. An automatic method for detection, measurements, and species classification of individual trees is outlined here. First, individual trees are detected using a canopy model produced from laser data. Second, crown area and tree height are derived for each tree. Third, classification is performed based on features derived from laser data for each detected tree. This method has been validated on twelve field plots (20×50 m) at the test site Remningstorp in southern Sweden. The field plots were dominated by Norway spruce (Picea abies L. Karst.) and Scots Pine (Pinus Sylvestris L.). In total, 71 % of all trees with a stem diameter >0.05 m were detected. Since a large portion of the undetected trees had a small diameter, 91 % of the total stem volume was detected. Height and crown diameter of the detected trees could be estimated with a root mean square error (RMSE) of 0.63 m and 0.61 m respectively. Stem diameter was estimated with an RMSE of 0.038 m using laser measured tree height and crown diameter. Thus, most visible trees could be detected using laser data and tree height could be measured with the same accuracy as conventional field measurements of individual trees. For classification a tree species classifier was trained with trees on all plots except for the plot that was classified. It was found that spruce could be separated from pine on an individual tree level with an overall accuracy of 95 %. In ongoing work another classifier have been developed and trained to separate between pine, spruces and deciduous trees. Work has also been performed on combining laser data and aerial imagery to support the classification process. Finally, we present some result on 3D visualization of forest. Laser data and aerial imagery are used in combination to build 3D models of forest landscapes

Publicerad i


Utgivare: Swedish University of Agricultural Sciences, Department of Forest Resource Management and Geomatics

Konferens

The ScandLaser Scientific Workshop on Airborne Laser Scanning of Forests