Schardt, Mathias
Figure 8: Crown maps (left) and segmentation - light lines - vs. crown maps - dark lines - (right)
clipping of Alpine test site
segmented trees / reference trees 17 / 19
mean of differences reference trees - laser-derived maxima | 0.42 m
standard deviation 1,17 m
Table 6: Statistics for laser-derived tree heights based on segmented tree height map
Further, test showed that the method works quite well for trees standing not to dense. Especially coniferous trees were
segmented correctly in most of the cases. In dense areas the method tends to merge tree crowns. The only sensible
parameter of the whole method is the smoothing scale. The optimal scale depends on the crown sizes of the trees in the
examined area. Especially when there exists a variety of crown sizes it is difficult to find a optimal scale. Due to this fact
future work will focus on adaptive blurring and scale space approaches to improve the smoothing part of the method.
Nevertheless, statistics for segmented trees concerning tree heights are better or at least within the accuracy of field-based
measurements assumed to be +0.5m, table 6.
6 CONCLUSION
Advantages of High and Very High Resolution Satellite Data
It is to be expected that the new very high resolution represents a major breakthrough, allowing the utilization of the above
discussed forest parameters derived from satellite images in regions characterized by small stand size, as it is typical in
most of the Austrian forests and especially in the test sites to examined in the proposed investigation. This is especially
true for inventories in forest areas, which are managed in an natural manner. Thus, by using very high resolution satellite
data the weaknesses of Thematic Mapper and SPOT data listed under section 1 and 2 can be surmounted.
In spite of the advantages of very high resolution satellite data it is not to be expected that the parameters listed under
table 2 which are of predominant importance for forest inventories can be assessed. This is due the fact that it is not
possible to derive 3 - dimensional information out of the signature of the satellite images. Considering the state of the
art concerning digital photogrammetry and satellite stereoscopy the extraction of precise 3D - forest information from
existing or future satellite systems seems not to be feasible.
Advantages of laser-scanner data
Forest inventories based solely on laser-scanner data are not feasible since parameters such as tree species types and age
classes cannot be extracted from these data. Nevertheless these data can be used to compensate the general drawbacks
of satellite imagery by providing supplementary information on parameters that cannot be classified by means of satellite
remote sensing. Thus, synergies resulting from the usage of both sensor types will significantly increase the performance
of remote sensing based forest inventories. The potential of laser-scanner data for the assessment of forest inventory data
has already been proved by several investigations (Nässet, 1997, Hoss, 1997).
REFERENCES
Hoss, H., 1997. Einsatz der laserscanner-verfahrens beim aufbau des digitalen gelándemodells (dgm) in baden-
württemberg. Fachzeitschrift "Photogrammetrie, Fernerkundung und Geoinformation".
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 1323