Sander Oude Elberink
trees in the laser scanner image,
as one can see in the FLI-MAP
data set in figure 6. Figure 7
and 8 show the 1-meter
resolution Optech data set. One
can see the extraction. of
buildings and trees extraction
on the hand of texture measures
in figure 7 or the (more
satisfying) extraction using
simultaneous first and last
pulse in figure 8. In data sets
with greater than 1.5 meter grid resolution the result of the contrast texture measure does not lead to a satisfactory
distinction between buildings and trees anymore. :
Fig 8: Left: last pulse DSM; center: trees extraction by first minus last pulse; right;
building extraction by morphological filtering of last pulse DSM at non-tree pixels,
3.4 Flowchart
A
fa I 1
Raw laser DSM ı DSM : Reflectance
scanner data First pulse Last pulse ! image
t
;
! t
i t
t i
1 Trees Normalised ' Trees : Roads
£x ces * % 4 3
mace Processing: detection by DSM by : detection by ! detection by
filtering texture morph. | first minus. ! texture
measures filtering last pulse measures
lm mmo ommo -
Non-tree pixels
4
z
,
4
,
4
4
,l
* v
>
; : S. i Classification
Classification Non-classified pixels of roads, grass
: of buildings, and apric
Image processing: Hess sheds ———————— J and agricultural
classification E fields.
Fig. 9: Flowchart from data to classification.
Figure 9 shows the flowchart of the raw laser scanner data towards the classification result. Because the registration of
last pulse data is not necessary, but desirable, and the fact that only a few systems are able to register first and last pulse
simultaneously, the flowchart of the last pulse data is shown by a dotted line. The reflectance value, while it is not
recorded by all systems, is a crucial pre-requisite for the classification of roads, grassland and agricultural land.
682 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.
press 1
Pept a. pe Pes
. T t5 rT
