ISPRS Commission III, Vol.34, Part 3A „Photogrammetric Computer Vision“, Graz, 2002
If this is done correctly, the overall model test of the least
squares strip adjustment (F-test) will be approximately equal
to 1, indicating an appropiate model
A fully filled covariance matrix for the observations implies a
much longer processing time for the strip adjustment. From
several datasets we learned that the resulting standard
deviations of the strip offsets are about a factor 1.3 higher
when taking into account the off-diagonal elements. In order
to save time, the adjustment is therefore done with a diagonal
covariance matrix for the observations, taking into account
error component 2 and the sill of error component 3, but
neglecting the off-diagonal elements (correlations). To
correct for this, the resulting strip offset standard deviations
are multiplied by a factor of 1.3.
In figure 6 strip offset standard deviations are visualised per
strip. It is apparent that the standard deviations of the offsets
are smaller in areas with relatively more ground control
points, see dots, (and also more tie points, not visible in
figure 6) and in areas with more cross strips. The right part of
the block has 3 cross strips in opposition to the left strips
with only a single cross strip.
140000 \
N au
150000 ^N a ee et
ME M 4000050000560) ;
1 00005200005300005
RD x (F%% 450000470000480000490000500006%
RD y [m]
Figure 6. Standard deviations of strip offsets.
From the standard deviations of the strip offsets, a single
standard deviation for the precision of a complete laser
dataset (entire block) can be calculated by applying the
propagation law of variances:
i=n j^!
N
2
26; * 2,000)
i=1
2 J=1 (4)
O dataset —
n:n
with:
iandj = numbers of adjacent strips
e — standard deviation of offset a of strip i
Oa, = covariance between offsets of strip i and strip j
n — total number of strips in the laser dataset
5. RESULTS OF REAL DATA ANALYSIS
The Dutch Survey Department is in the fortunate position to
have laser altimetry data available for almost the complete
country. Therefore the amplitude of the different error
components can be calculated for large and numerous data
sets (approximately the size of a province). Table 1 gives
typical values for the amplitudes of the different error
components. Note that outliers occur in practice.
error per mean max dimens. |
1 | point 0.08 0.10 m
2 | GPS observation: nugget | 0.03 0.05 m
3a | strip: Jsill 0.04 0.05 m
3b | strip: range 9 15 km
4a | block: Ostrip offsets 0.03 0.08 m
4b | block: strip offsets 0.03 0.10 m
Table 1. Typical values for the error components
6. UTILISATION OF ERROR DESCRIPTION
The new height error description will be used for two
purposes. The new quality demands for laser DEMs,
delivered by laser scanning companies, are based on this
description. On the other hand, the customers (DEM users)
will be provided with this extended quality description.
6.1 Towards the laser scanning companies
The Dutch Survey Department intends to use the new error
model with suitable amplitude requirements in the contracts
with the laser scanning companies. The calculation methods
described in section 4 enable fast and (nearly) automatic tools
for the assessing of the data delivered in strips from the
contractors.
Table 2 shows the maximal allowed error amplitudes per
error type. For every error component the mean value and the
maximal occurring value (max) for an entire dataset (project
area) has to be below these thresholds. Note that the
demanded amplitude for error component 4b (strip offsets) is
quite strict. The reason is that this error (systematic per strip)
can easily be corrected by strip adjustment. We expect that
the companies already have executed a strip adjustment.
error per mean max dimens.
1 | point 0.12 0.24 m
2 | GPS observation: {nugget | 0.03 0.06 m
3a | strip: sill 0.05 0.08 m
3b | strip: range 10 30 km
4a | block: Ostrip offsets 0.05 0.08 m
4b | block: strip offsets 0.05 0.13 m
Table 2. Contract demands for maximal error amplitudes
6.2 Towards the users
Moreover, the new height error model will be used for the
description of the precision of laser altimetry deliveries to
customers. With the given error amplitudes per laser dataset,
customers (or DEM users) are able to compute the precision
of derived products from the laser data, e.g. volumes or mean