International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
LIOAR Data Points
“Original UDAR Data Forts
Collected Ground Points u
omoving the height bias
Northing +575000 (m)
5
Heigth (ny
2 a si 528 230 s
Easting +913000 (m) Easting «913000 (m)
(a) (b)
. , na ;
/ Original LIDAR Dats Points i
20% LIDAR Dat noving the height bias.
53:1 —— LIDAR Data Points after the shift ín height and easfing |
0.085 —— Collected Ground Points i
2 !
oor {
ss i
0%: = i
Eos £ ]
5 = L i
Ty 4
m 2 |
oh — TV behavior with different shâts in casting :
+ _ Stift in X staistying the best match _ e
1 5 2 3a 5%
“shi dm S rs ep xs EUM
(c) (d)
Figure 8: Planimetric offset in (X) direction at location EW3R.
As stated above, eight locations were tested to obtain the
planimetric shift. In fact the number of these locations was
limited since the area that was covered by the topographical
survey does not contain many suitable features for that purpose.
Table 1 summarizes the results at those selected locations.
Regarding the offset in X (Easting) direction, which is across
the flight direction and coincide with the scanning direction, six
locations were selected, three at the edge of the swath width of
strip 2 (EWIR, EW3R, and EWSR) and the rest at the middle
of the strip (EWIL, EW3L, and EWSL). As expected at the
strip edge, the offset in the scanning direction (-0.60m as an
average) was larger than at the middle of the strip (-0.30m as an
average). However, those shifts were in the same direction. The
same thing could be said for the height bias, height offsets seem
to be larger in magnitude at the edge of the strip. Unfortunately
there were no significant features at the other end of the strip to
have a complete idea of the planimetric accuracy behavior
along the whole swath width. On the other hand, two locations
were selected to test the accuracy along the flight direction Y
(Northing), one at the middle of the strip (NS1) and the other
one at the edge (NS2). The two locations show an offset of —
0.55m and —0.40m, respectively, in the same direction.
: Height pim Plan. Location in
Location iD Direction Offset swath
bias(m) :
(m) width
EWIR -0.12 Easting X -0.67 Right edge
EWIL -0.03 Easting X -0.28 Middle
EW3R -0.23 Easting X -0.54 Right edge
EW3L -0.07 Easting X -0.20 Middle
EWSR -0.23 Easting X -0.62 Right edge
EWSL -0.08 Easting X -0.43 Middle
NSI -0.04 Northing Y | -0.55 Middle
NS2 -0.07 Northing Y | -0.40 Middle
Table 1: Planimetric accuracy results.
6. RESULTS, ANALYSIS, AND STATISTICAL
CONCLUSIONS
To show the error behavior with respect to time, the LIDAR
data points over the test area were sorted based on the time they
were scanned. Then the calculated height differences (1008
biases) were sorted accordingly. The test area covered only less
148
than two seconds of the scanning time which contains about
19,000 data points. Although the size of the test sample (1008
points) is sufficient to represent the sample, the distribution of
the test points across the time span (the two seconds) was not
ideal. Figure 9 shows the behavior of the absolute height error
with respect to the time they were scanned.
e T 1 MER rs T a
| Calculated differnces |
Short period variance
0.2H Estimated trend
o
>
e
1
o
-
Height diff. (Ground - LIDAR) (m)
o
o
w
— — — — —À— -— SO OS
0.2 0.4 0.6 0.8 1 12 14 1.6 18 2
Time (sec)
Figure 9: Height accuracy (Ground — LIDAR) versus time.
The height differences between the LIDAR points and their
corresponding ground-surveyed points show two types of
variations. The first type of variation is called short period
variation. This variation has a high frequency as we can see in
figure 10. This variation between two consecutive points could
reach 0.30m as a maximum within 0.001sec. This short
variation of the uncertainty of LIDAR heights gives an
indication of the system precision since the consecutive points
are so close to each other in the time domain and the test was
conducted on a flat surface where the height is very nearly the
same. On the other hand, as shown in figure 9, a data driven
trend of the differences is observed which is the second form of
the variation. A “trend” is defined in (Mikhail, 1976) as “it is
that component of a random phenomenon which has a period
larger than the recorded data sample”, which can be seen
clearly in figure 9. Although the test data represent only a small
sample, this trend is very noticeable. The planimetric relative
accuracy between data strips, see figure 5, ascertain this
conclusion regarding the two forms of the random variation
since the computation of the relative accuracy covers most of
the data.
| - Calculated differnces |
| Short period variance |
1 1 1 1 1 i i
0.79 0.795 0.8 0.805 0.81 0.815 0.82
Time (sec)
Figure 10: Short period variation of the height biases.
To conclude, the results will be summarized. A height offset of
0.08m was found between the surveyed ground points and the
LIDAR data in the test area. The computed uncertainty of the
Internat
LIDAR
slightly
the syst
comput
show tv
larger (
middle.
based o
swath v
middle
whole s
the com
Both a
variatio
variatio
variatio
system.
“a trenc
be as a
is mode
data w
differen
these cc
The au
researct
Office z
Baltsav
systems
164-196
Burmar
correcti
Photogr
pp.125-
Casella,
accurac
exploiti
Amster
Georgia
multipa
Geodae
Gonzak
New C
Mappin
pp. 67-7
Grewal
System,
Sons, Ir
9
Mikhail
squares
2481-5.
Mikhail
2001. 7i
Sons, lr
