45m
»
167 points
Figure 2. Test site
878 points
Figure 3. Intensity image
3.1 Noise Reduction
In the case of ground based laser scanning, many noises which
were caused by trees are estimated, and these noises have to
remove by automatically. Hence, automatic noise reduction
method was investigated.
Detail procedures for noise reduction are as follows, and Figure
4 shows the flow of automatic noise reduction method, and
Figure 5 shows position of topographic and noise data.
Three procedures were investigated as noise reduction methods,
the first is noise reduction by Z values, the second is performed
using distance between neighboring points, and intensity image
was used at the final step.
eNoise Reduction by Z values.
+ Search a point which have smaller Z values and have similar
XY values compare with an interest point in neighboring an
interest point.
+ If detect such a point, an interest point is removed as a noise.
+ If an interest point has the minimum Z values, an interest
point is identified as a ground.
Repeating the above procedures, noises which caused by trees
were considerably removed.
However, there are still some noises in mass area of trees. In
order resolve the second issue, the following procedures were
investigated.
+ Select an interest point in the mass area using intensity image.
+ Calculate distances between an interest point and neighboring
points.
+ If there are many shorter distances, an interest point is
identified as a ground.
+ If there are many longer distance, an interest point is reduced
as a noise.
As the final step, a few noises were reduced as a labeling
procedure using intensity image.
Aen)
| Pick a Interes Point
TE T
No d =
——— Same XY Value
—
—
s?
S qe e
y Yes
< Smaller Z Values?
Sma u
No
|
|
|
|
LS NO
»— Shoter Distances? 7
M uu
S es .
y Yes | Noise |
| Topographic Data |
a er ed
Figure 4. Flow of automatic noise reduction
A
ZA
Z3 Noise data
c — pem m ee
—»| Topographic data
Figure 5. Position of topographic and noise data
-96—
ca
se:
sit
po
