nbul 2004
spectively
ENT
s requires
luation of
arameters.
e a valid
ies follow
nalysis of
Iscrepancy
individual
Z.Ox are
nce level,
eviation of
wo
Cn
S
gat |,
B,C,
,
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
C
2 sige " T. 3 sigma
discrepancy in y-direction( Yref - Yc)
1 sigma A i ^ e
D
2 sigma n i prim 3 sigms
discrepancy in y-direction(Yref - Yc)
discrepancy in x«direction (Xref - Xc)
o ; AT o ; silicon
Planimetric discrepancy ~ Mean Planimetric discrepancy
Figure 3 Discrepancy scatter plots, for time-averaged positions,
A) 1 minute averaging, B) 3 minutes averaging, C) 5 minutes
averaging, D) 10 minutes averaging.
Practical discrepancy outliers at lo, 20, and 30, for 1, 3, 5, and
10 minutes are presented in table 7
Percentage of outliers
Practical%
Ze Ee 3 Im 3m 5m 10m
S 3 NY XN X.Y x Y
2 E
&
6827 31.73% ARE 4023 35310 305
2 95.45 455 9 «q sq $y 99
99.73 0.27 0 41 0 3 0 0 0 2
Table 7. Practical discrepancy outliers for 1, 3, 5, and 10
minutes
From this table we conclude that all Y discrepancies are
satisfying the test of a, against H, , at 68% confidence and
for all 4 different averaging time, but in X direction the number
of practical outliers are larger than the theoretical outliers for all
averaging times. At 95.45% confidence all the practical outliers
are larger than the theoretical outliers, except the X outlier at 3
minutes averaging time. At 99.97% confidence all
631
discrepancies in X direction are satisfying the test of Hy
against H. in Y direction they are all larger than the
theoretical outliers, except for 5 minutes averaging time.
6. DISCUSSION
When reference points are available the user can perform code
measurements to estimate the bias and random error, and use
the results as a figure of merit for the measurements in the area
of interest. The reference points may be obtained from national
mapping authorities or from precise carrier-phase instruments
measuring in differential mode. Preferably the coordinates of
the references points should be defined in WGS84 to eliminate
the effects of differences in datum transformation parameters as
applied by the mapping authorities and those used by the code
receiver handheld units. Averaging can reduce the effect of
random error and provides an indication of the systematic
errors. The results of our averaging experiment show that the
standard deviation is reduced with 65 percent while the bias is
reduced with 47 percent from | to 10 minutes averaging. If this
difference in the reduction of the accuracy and precision would
be significant, it implies that the EPE reports become better
predictors for the accuracy with increasing averaging time. This
initial result, however, needs to be confirmed and we
recommend users to repeat measurements on different hours of
the day and monitor the effects of variations of constellation. In
addition, one may refer to HDOP values, which may be
provided by some instruments. The mean, RMSE of HDOP,
and the mean number of available satellites can be used to
compute an average for the expected errors. In table 8 these
parameters are given for the present experiment.
HDOP reported by handheld code receiver
Mean HDOP 2.23
RMS HDOP 0.31
Mean no. Satellite — 6
Table 8. Number of satellites, HDOP values and RMS of
HDOP for handheld code receiver.
If no reference points are available, the averaged position is
often used as an approximation to the actual position (Wilson,
2002). This will generally lead to an underestimation of the
predicted error because the bias is ignored. It is important,
therefore to convey this to the user and to carefully assess the
range in systematic errors in the accuracy assessment
experiments of a handheld unit. Alternatively some units
provide statements on probable errors, such as the horizontal
estimated position error (EPE) or figure or merit (FOM) for
average positions (DePriest, 2002). FOM estimates are usually
different for each company and computation is held
confidential so that there are no means available to evaluate
their robustness.
Regardless reference points are available or not, the user of
handheld units currently lacks information that allows one to
