its mean square error
; V6 ii, ;
no z t pm —g ; in which are (4)
JI] an (5)
m
Oo
[vv] = ze - d), (6)
The standard error of unit weight m, (with the dimension
density value) has to be considered as stochastic quantity
only, if the linear dependency of two spectral channels and
thus the selected functional model fits. For this, Mo will
achieve its minimum in the magnitude of the accuracy to be
expected by the scanning system. In all other cases, the mo
will inerease due to its additional deterministic portions.
5. Sample Computations
The disposition as described in section 2 has been applied
to li-spectral data of the LANDSAT satellite. As it had to be
| expected that different results of the adjustment would be
| obtained depending on the topographic features (type of land-
| scape) and their measured spectral data, different areas
were selected from a full scene comprising 2340 scan lines
with 3240 pixels each:
Area 1: Minden, 240 lines with 360 elements each
Area 2: Lehrte, 240 lines with 360 elements each
Area 3: North Germany, southern part
790 lines with 2040 elements each.
The available spectral ranges have to be assigned to the
LANDSAT channels 4, 5, 6, 7 (= Kl, K5, K6, KT). The follow-
ing channel combinations were formed: K5 = F(K4), K6 = F(K4),
K7 s f(KÀ), K6 = F(K5), K7 = f(K5), KT - f(KR6).
The results of the adjustment and their mean square errors
are listed in Table 1. (There, M is substituting the arith-
metic mean.) The smallest mg-values result for the following