期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2004
卷号：XXXVI-8/W2
出版社：Copernicus Publications
摘要：The purpose of this study was to test a method for delineating individual tree crowns using a fully automated object-based pattern recognition methodology. The study material included small-footprint time-of-flight laser scanner data acquired over the Norway spruce (Picea abies) and European beech (Fagus sylvatica) dominated forests of the Bavarian Forest National Park near Passau, Germany. Ground reference data was collected in 28 sample plots, each 0.1 – 0.25 ha in area, containing a range of stand types. Laser data was acquired in the spring and summer of 2002 using a Toposys II airborne laser system. The concept of the evaluated delineation methodology is a pouring algorithm for the estimation of treetops and crown areas basing on a digital crown model. In order to enhance the results additional functions like ray tracing have been introduced to complete this main concept. Three individual tree attributes were sought in this analysis: stem position (x, y, z), tree height and mean crown radius. The number of trees that were detected using laser data was compared with field measurements. Altogether 2666 field measured trees were used for the accuracy assessment. The overall detection rate was 37.5 %. In the top layer 58.0 %, in the middle layer 5.2 % and in the lower layer 1.4 % of the trees were detected. The percentage of detected volume was 74.1 % of the total volume measured in the field. The tree detection was a function of stock density and the number of trees per hectare. The percentage of detected volume was better for coniferous trees (76.3 %) than for deciduous trees (70.4 %). High correlations were found between laser and ground measured tree height, with RMSE values of 1.44 m for all trees, of 1.38 m for coniferous trees and 1.51 m for deciduous trees. The correlation was also high between laser estimated and field measured crown radius. For trees within 20 stands, the RMSE was 0.94 m. The estimations for the coniferous tree crowns (RMSE 0.61 m) were more accurate than the measurements for the deciduous tree crowns (RMSE 0.96 m).
关键词：Forestry; Remote sensing; Laser Scanner; Individual Tree Detection
