' chosen
ppect fo
e optiori-
IC various
jo first
he the-
'elatively
“ically
Cc
)S in-
Lk
Y(X5, Y.)
ements
non-para-
jor orien-
ints,
) not
"esented
‘ation x,y
red data
molation
solved
MULTI TEMPORAL
MULTI-SENSOR IMAGERY
CARTOGRAPHIC
DATA BASES RADIONETRIC e| | e MSS IMAGERY
© AERIAL PHOTOGRAPHY
e TOPOGRAPHY e IR LINE SCANNER
GEOMETRIC € RADAR (SLAR,SAR)
CORRECTIONS
RANGE OF
9 BOUNDARTES DATA PROCESSING
e SOIL TYPES GEOMETRIC TRANSFORMATION
ie INTO THE DATA BASE SYSTEM
: RELEVANT
e
ANCILLARY DATA ORT
e GEOLOGY MULTI-SENSOR
SELECTION OF OBJECTSIGNATURES
e LAND USE TRAINING SITES AND -TEXTURES
| OBJECT RECOGNITION AND CLASSIFICATION |
i i
UPDATING OF D/A CONVERSION (FILMWRITE)
DATA BASES (E.G.LAND USE MAP)
Figure 1 Combined processing of data bases and integrated imageries
Finally, the interpolation of a point x,y is carried out by using the
equation
= n
7=2z+DK=-2+ EX k f(s.) (5)
i=1 J J
J
2.3 Multiquadric geometric processing
The determination of the geometric transformation functions X and Y
is performed in two steps.
First, trend functions Xi" X > Y) and Your Yt (X5, Y.) are derived
as twodimensional polynomials of low order or the sensor modelling function,
respectively. These trend functions eliminate global distortions, scale affi-
nities, translations and common rotations between the two coordinate systems.
Second, remaining differential discrepancies X E X7 X41Tk and
Y ke Y Ak. Yu at the pass points k-1,K are modelled analÿKicaily as
multiquadric equations. The final interpolated coordinates become then
Xi- X, * XT and Yi7 Yi! * Yir
15
| e iU CM — OE