registration of a drift angle of 10°; this is due to the row
geometry and the dimensions of the lens. Problems in ascertain-
ing the MAD minimum arise if disturbing factors prevent the
formation of a function "trough" and the MAD function dege-
nerates into an inclined strairht line. In that case the al-
gorithm finds a minimum at the boundary, which of course has
to be discarded.
MEASURES TO ELEMINATE DISTURBANCE
Prom the previous passages it is evident that the state vari-
bles n and ,u are subjected to a variety of disturbing influ-
ences. It sÜegestc ltself, therefore, that the values ascer-
tained should be filtered before feeding them to the control
elements, Since it is posable to define a linear, though simple,
model of the aircraft’s motion, to classify the disturbances
with regard to their points of attack and to estimate their
standard deviations, an optimum linear filter (KALMAN filter)
can be devised, if we assume independence between state vari-
ables and disturbances as well as independence of the distur-
bances from each other. Such a filter has a number of advanta-
ges which will be explained later.
The signal generation model corresponds to the system equa-
tions of the system "aireraft-camera", with n(t) being the
only state variable present. If we assume that in flight along
a route at a constant height and with preset wind correction
angle and speed, n(t) must be constant, then changes of n(t)
can exclusively caused by disturbances (turbulences, thermal
currents, cross-wind swaying). Hence, we arrive at a very
simple motion equation:
n(t) » b. f (t) (3)
where f (t) stands for system noise, which may be assumed to
be white noise. Further, the observation equation of the pro-
cess has to be formulated:
Y3 = ny + N, (4)
in which N. designates discretey white disturbances of the
measurements with a Gaussian distribution, and y. stands for
the n(t) values observed, i.e. obtained by the measuring
procedure described above. From the above assumptions, the
standard deviations of the disturbing signals can be stated
to be
Bhf (6) fls1sY = B= 125 = 8) (5)
E(N,N,) = 5, 9433) (6)
with 6" desicnating Kronecker's delta.
131
