In: Wagner W„ Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
LIDAR data, branches of 27 coniferous and 38 deciduous trees 
were derived by calculating the mean silhouette values 
repeatedly for different ¿ values using simple ¿-means approach 
for improved visualization. The method lacks the efficiency for 
finding the suitable value of k with respect to different tree 
types, tree age and forest conditions. In their study the result is 
not validated using any field data. 0rka et al. (2009) tested the 
supervised classification strategy using linear discriminant 
analysis (LDA), random forest (RF) algorithm and support 
vector machines (SVM) for tree species classification. They also 
used unsupervised ¿-means clustering and ¿-means clustering in 
combination with the unsupervised random forest algorithm for 
the same purpose. However, their result showed that accuracies 
were lower in case of unsupervised one than for supervised 
methods applied for overall species classification. This shows 
that supervised methods are more promising which was found 
true during the investigation after a comparative qualitative 
analysis of the output by applying different clustering 
algorithms (Gupta et al., 2010). Vauhkonen et al. (2009) used 
LDA for the classification of individual trees and alpha shape 
metrics for tree species classification in Scandinavian test site 
comprising 92 trees detected and delineated manually from a 
very dense ALS data. However, the method applied required to 
be tested for larger dataset with lower or medium point density 
by automatic detection. Apart from tree detection methods, a 
method for reconstructing the tree crowns was also provided 
(Pyysalo and Hyyppä, 2002). There are many ways for 
constructing the shape of extracted points of single tree using 
different computational geometry concept like convex hull, 3D 
Delaunay triangulation or can be shown as 3-D surface or mesh. 
3. MATERIALS AND METHODOLOGY 
3.1 Study Area 
Investigations were carried out in the selected plots of 
administrative forest district Hardt, Baden-Württemberg region 
in the South-West of Germany. The study area is flat, 1.75 ha in 
total, comprising 7 rectangular study plots, each of size 0.25 ha. 
Figure 1. Location of rectangular study plots (green) as seen in 
the RGB aerial photograph 
The study plots in the forest are characterized by a variety of 
deciduous and coniferous tree species of different ages. The 
forest is marked with highly interconnected and dense standing 
deciduous crowns. The overall fraction of deciduous and 
evergreen coniferous tree species is nearly 82% and 18%, 
respectively. Except plot 3, rest of the plots are dominated by 
deciduous trees species (Table 2). The dominant tree types in 
the studied field plots are Scots pine (Pinus sylvestris - 13.9%). 
Cherry (Primus avium - 23.7%), Oak (Quercus petraea ~ 
24.3%), European Beech (Fagus sylvatica - 21.1%), Hornbeam 
(Carpinus betulus - 9.5%) and 7.6% others species (Norway 
spruce - Picea abies, Douglas fir - Pseudotsuga menziesii and 
few other minor species). All the study plots are made up of 
multi-storey canopy layers. From the field inventory data 
collected, it was found that the height of different tree species of 
analyzed study plots varied between 8-35 m and average height 
ranged between 8-31 m. 
Plot 
Id 
Tree type 
Tree species 
% of tree 
1 
Deciduous 
Hornbeam + Cherry 
34.4 + 56.2 
= 90.6 
Evergreen 
conifer 
Scots pine 
9.4 
2 
Deciduous 
Cherry + Oak 
91.8+4.1 
= 95.9 
Evergreen 
conifer 
Scots pine 
4.1 
3 
Deciduous 
Red Oak + European 
Beech + Black 
Locust 
8.3+25.0 + 
5.6 = 38.9 
Evergreen 
conifer 
Scots pine + 
Douglas fir + 
Norway spruce 
44.4+13.9 
+ 2.8 = 61.1 
4 
Deciduous 
Hornbeam + 
European Beech 
68.2 + 9.1 
= 77.3 
Evergreen 
conifer 
Scots pine 
22.7 
5 
Deciduous 
Oak + European 
Beech + Linden + 
Silver Birch 
73.2+18.8 + 
3.0+ 1.0 = 
96.0 
Evergreen 
conifer 
Scots pine 
4.0 
6 
Deciduous 
European Beech + 
Sycamore Maple 
72.4 + 2.1 
= 74.5 
Evergreen 
conifer 
Scots pine 
25.5 
7 
Deciduous 
European Beech + 
Sycamore Maple + 
Cherry + Hornbeam 
+ Oak 
10.0 + 6.7 + 
40.0+13.3 + 
3.3 =73.3 
Evergreen 
conifer 
Scots pine + Norway 
spruce 
6.7 + 20.0 
= 26.7 
Table 2. Tree species distribution 
The name and distribution of tree species in Table 2 are in the 
same order. 
3.2 Field Data Characteristics 
Forest inventory data of the study plots was provided by the 
Forest Research Institute (FVA) of Baden-Württemberg. All 
trees in the plot above 7 cm diameter at breast height (DBH) 
were measured. Two top heights of the main tree and one top 
height of the dominated tree were measured using a Vertex® 
instrument. The arithmetic mean of the height measurements 
was calculated as an average top height for each plot (Straub et 
al., 2009). Stand height curves with the DBH as input variable 
were used to estimate the heights of the remaining trees (Kom- 
Allan et al., 2004). Several height percentiles were calculated