UPDATING THE CONTENT OF A GIS DATABASE BY THE ROBUST FUSION
OF DIFFERENT LOW COST GEOMETRIC SOURCES
IN GENERALIZED ANALYTICAL MIXED MODELS
Fabio Crosilla*, Ambrogio Manzino**, Domenico Visintini*
*Department of Civil Engineering, University of Udine, Italy
**Department of Georesources and Territory, Politechnic of Torino, Italy
ISPRS Commission IV, Working Group 3
KEY WORDS: Cartography, Updating, Algorithms, Low Cost Sources, Analytical Robust Fusion.
ABSTRACT
The paper proposes an original method for the process of simultaneously georeferencing the digital version of several cartographic
units, belonging either to the same or to different map projections, and updating the same cartographic product by the restitution of
the particulars taken on the ground with amateur or digital photogrammetric sensors.
The georeferencing procedure is performed by a classical 8-parameter transformation with the introduction of constraint equations
for the tie points of contiguous map units, while the updating is carried out by Direct Lirear Transformation (DLT).
Both analytical processes are joined in a common mixed model of estimation and prediction. The coordinates of the points to be
updated are submitted to a BLUU estimate, together with the orientation parameters of the images, the cartographic transformation
coefficients and the tie point coordinates of cartographic and photographic layers, used to orient the pictures.
The coordinates of the cartographic control points, possibly seen on some pictures, are predicted.
Some numerical experiments of updating the digital version of a technical map are finally reported.
1. INTRODUCTION
The wide spread of GIS in the last years requires the availabi-
lity of reliable and up-to-date digital mappings, at different
scales.
Although the best and completest cartographic products can be
today obtained by photogrammetric survey at a proper scale, a
correct scanning of existing cartography submitted to a robust
georeferencing process, can give acceptable cartographic
products for GIS applications.
For these reasons, the kind of updating and testing procedures
to adopt become prominent. These must be highly productive,
of low cost and easy realization, of sufficient precision and, just
for the testing, able to verify the accuracy of global cartographic
entities and not only of isolated points (static or pseudo-static
mode) or lines and trajectories (kinematics mode) as can be
done for both methods with GPS techniques.
The use of amateur or digital sensors for updating the map with
the necessary particulars by large scale pictures taken on the
ground, is a technique that can dramatically reduce the costs
and improve the productivity, if point coordinates used to orient
the images are extracted from an already existing digital map or
from a digital version, obtained by scanning procedure, of a
map.
In this case a careful analysis of error sources characterizing
non metric images and the point coordinates used to orient them
must be done, and a specific mathematical model able to
predict such kinds of error, must be applied.
The paper proposes an original method for updating a digital
map and testing its quality, by the analytical fusion, in a mixed
model of estimation and prediction, of the georeferencing
process of several kinds of digital maps, with the object
reconstruction of the necessary parts, obtained by Direct Linear
Transformation (DLT) of non metric pictures.
Among all possible estimations and predictions, the paper
considers the ones satisfying optimality conditions in terms of
"mean square error", that must be the minimum among all
possible linear estimators and predictors, and in terms of
"uniform umbiasedness", that must be valid not only for the
194
true but unknown values of the terms to be estimated or pre-
dicted, but for any value of them.
The coordinates of cartographic points to be updated and tie
points, together with all transformation parameters, are
submitted to a BLUU estimate, while cartographic control
points that can be seen on some pictures are predicted.
For control points, GPS measurements allow to fix a correct
datum to the digital mapping, since the transformation from the
GPS reference system to the cartographic one is precise enough
for small areas.
The computation of a multiple georeferencing procedure and
the simultaneous solution of a DLT algorithm for updating the
contents of the map layers allows to consider the correlation
existing between the different data acquisition techniques, and,
at the same time, makes the computation procedures more
robust and reliable than single transformations.
Particular emphasis is given to the correct weighting of the
functional model relating to each transformation. In this regard,
a robust method of assigning the observational weights accor-
ding to the residual values of the observational equations is
proposed (Krarup, Juhl and Kubik, 1980).
Some numerical experiments relating to the updating of the
digital version of the cartographic product of the Friuli-Venezia
Giulia regional administration at the scale 1:5.000, put in
evidence the capability and the limits of the proposed method.
2. ANALYTICAL MODEL
Lets take into account an 8-parameter transformation 3$
analytical model for simultaneously georeferencing the digital
version of contiguous cartographic units or different map layers
obtained by scanning procedures:
ri aX, + bY; t|
1 uX,+vY, +1
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
whe
Xi ,
refer
Xi,
refer
aj A
para
Tos
logo
cont
its fi
whei
para
Ay
grap
grap
X1 ]
coef
X 4 i
coor
sis €
Tor
by p
calle
whe
refe
Als
(2)1
whe