In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
603 
the laser and aerial image features extracted from grid elements 
and automatically delineated stand polygons in order to find out 
which unit serves better the purpose of extracting image 
features for estimating forest attributes. 
2. MATERIAL AND METHODS 
2.1 Study areas 
The laser scanning and aerial image based estimation was tested 
in two study areas. Study area 1 was located in the municipality 
of Lammi in southern Finland and its field data consisted of 282 
fixed-radius (9.77 m) circular field sample plots that were 
measured in 2007. Study area 2 was located in eastern Finland 
and 546 fixed radius (9 m) sample plots measured in 2009 were 
available as field reference data here. For the field sampling 
both study areas were stratified on the basis of earlier stand 
inventory data and the field sample plots were allocated to these 
strata in order to cover all types of forest in the study areas. 
There was some variation between the forest characteristics of 
the two study areas. In study area I the total growing stock was 
more evenly distributed between tree species groups pine, 
spruce and deciduous trees, whereas the study area 2 was 
clearly dominated by spruce. Furthermore, the study area 2 had 
somewhat higher average stand volume, as well as larger 
distribution of sample plot volumes. The statistics of the study 
areas based on the sample plots are presented in table 1. 
Study area 1 
Study area 2 
Average 
Max 
Average 
Max 
Total volume, 
m3/ha 
178.7 
575.4 
205.5 
798.5 
Volume of pine, 
m3/ha 
69.8 
560.6 
52.7 
561.8 
Volume of spruce, 
m3/ha 
63.7 
575.4 
109.3 
739.2 
Volume of 
deciduous, m3/ha 
45.2 
312.0 
43.5 
400.4 
Height, m 
17.0 
30.5 
17.3 
35.2 
Diameter, cm 
21.1 
50.2 
21.9 
60.3 
Table 1. Forest statistics of the study areas (average and 
maximum values of sample plots) 
2.2 Remote sensing data 
In study area 1 the remote sensing data consisted of color- 
infrared digital aerial imagery (containing near-infrared, red and 
green bands) and ALS data acquired from a flying altitude of 
1900 m with the density of 1.8 returned pulses per square meter. 
In study area 2 the remote sensing data consisted of color- 
infrared (containing near-infrared, red and green bands) and 
natural color (red, green and blue bands) digital aerial imagery 
and ALS data acquired from a flying altitude of 2000 m with the 
density of 0.6 returned pulses per square meter. Here the aerial 
image data was combined to a 4-band composite image 
containing blue, green, red and near-infrared bands. 
The aerial images were ortho-rectified and resampled to a 
spatial resolution of 0.5 m. The ALS point data was also 
interpolated to a raster image format (height and intensity 
images) using second degree polynomial model. The output 
laser images had similar spatial resolution as the aerial images. 
2.3 Automatic image segmentation 
Automatic stand delineation was carried out in the study areas 
by automatic segmentation of aerial images and ALS data 
interpolated to raster format. The segmentation was carried out 
in two phases. In the first phase initial segmentation was done 
using a modified implementation of Narendra & Goldberg 
(1980) algorithm, which employs local edge gradient. This 
method typically produces a large number of small polygons, 
and the objective is to find all potential segment borders at this 
phase. In the second phase the initial segments were processed 
using a region merging algorithm that was guided by parameters 
such as desired minimum size of final segments and the 
similarity/dissimilarity of the segments to be merged (t-ratio 
threshold). 
In this study, the initial segmentation was based entirely on 
laser height corresponding mainly to the stand height 
(Mustonen et al., 2008). The merging of initial segments into 
larger spatial units (i.e. final segments) was carried out on the 
basis of laser and aerial image data, taking into account also the 
tree species structure of the initial segments. Two automatic 
segmentations with minimum segment sizes of 350 m 2 and 0.1 
ha were carried out in both study areas. 
2.4 Extraction of laser and aerial image features 
Three remote sensing feature data sets were extracted from each 
of the study areas. In these sets the remote sensing features were 
allocated to each sample plot from a square window or a 
segment in which the sample plot was located. The feature set 
Grid was extracted from a 20 x 20 meter square window 
centered around each sample plot. The feature set Seg350 was 
extracted from image segments, whose minimum size was set as 
350 m 2 . The feature set SeglOOO was extracted from image 
segments, whose minimum size was set as 0.1 ha. 
The following statistical and textural features were extracted 
from the remotely sensed material for each feature data set: 
• Means, standard deviations and Haralick textural 
features (Haralick et al. 1973, Haralick 1979) of 
spectral values of aerial photographs, ALS height and 
intensity (first pulse only) 
• Height statistics for the first and last pulses of the 
points inside the field plot area or the segment area 
(Suvanto et al., 2005). These included mean, standard 
deviation (std), maximum, coefficient of variation, 
heights where certain percentages of points had 
accumulated and percentages of points accumulated at 
certain relative heights. Only points over 2 m in 
height were considered and the percentage of points 
over 2 m in height was included as a variable. 
• A number of std’s extracted from a 32 x 32 pixel 
window using block sizes from 1 to 8 pixels. 
All features were standardized to a mean of 0 and std of 1.