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：2005
卷号：XXXVI-3/W19
页码：144-149
出版社：Copernicus Publications
摘要：The objective of this paper is to improve the cost-effectiveness (accuracy of estimates versus applied costs) of present ITC (Individual Tree Crown) approaches which are based on 1) aerial imagery or 2) high density laser scanning data by introducing a hybrid technique (giving the height of each crown with laser data to image derived crown segment) and to compare accuracy of these three approaches. It is also studied whether the inability of ITC approach to separate tree crowns in dense forests affect remarkably on volume estimation. 78 carefully measured pine trees in 10 plots were used in this study. CIR image orthorectified with laser DSM was used as a major data source. Laser data were obtained from 400 m and 800 m (above ground level, AGL) acquisitions with Toposys I providing pulse density of approximately 10 and 5 pulses per m 2 . Results of current study demonstrated that the height of individual trees is the most important geometrical parameter for stem volume estimation that remote sensing methods can provide. Aerial image ITC approach, which has been lately even used in commercial operations, was significantly improved by including the height of tree from the laser data (individual tree stem volume R 2 changed from 0.14 to 0.54). This hybrid approach should be further studied as a potential operational forest inventory method, since it may be possible to derive ITC based solution using nationally acquired laser point clouds and already existing CIR imagery giving high quality results at stand level. The results indicate that individual tree volumes can be obtained with random error between 25 to 30 % and that volume related to small tree groups an be obtained with random error between 34 to 40 %. Both accuracies are relatively good, since the present Scots Pine models based on diameter (and requiring manual measurements for each tree) give accuracy of 17 %. The accuracies obtained for individual trees and small tree groups indicate that random error of less than 5 % could be obtained in future at stand level, however, the trees that cannot been seen from the remote sensed data needs first to be estimated by other means. This paper also demonstrates that even though remote sensing can not always provide ITC solutions, the stem volume estimates do not deteriorate significantly from this fact since the existing formulas take into account the widening of the tree crowns. From practical forestry point of view, it may not be so relevant whether all trees are correctly isolated with segmentation. Of course, the correct segmentation stays as a challenging scientific task. More emphasis in the future studies should be placed that the size of segmented crown relates to natural crown size, and how the conversion of remote sensing-derived crown size can be converted into stem diameter correctly
关键词：Individual Tree Crown (ITC); Laser scanning; Tree height; Aerial imagery; Segmentation
