Geography Reference
In-Depth Information
with the transformations, each transformation requiring a different number of control
points.
In fact, many kinds of georeferencing methods exist, and they produce results that can be
qualitatively and quantitatively different. In particular, the georeferencing algorithms can be
grouped in two different classes: global and local transformations (Balletti 2006; Boutoura &
Livieratos, 2006). In a global transformation (conformal, affine, projective, generic order
polynomial) the unknown parameters are calculated for the whole area. On the other hand,
in a local transformation (finite elements, morphing) the unknown parameters are calculated
for a small area, defined by a small number of control points or close to each control point.
The best georeferencing method probably do not exist, because the choice of a specific
georeferencing method depends on the specific case (map characteristics, number of
available GCPs, etc.) and on the purpose which the georeferenced images will be used for.
By means of a georeferencing process, the native metric content of the map being
reproduced in the digital image, the historical map can be compared with the present
cartography or other cartographies (also coeval ancient maps) in the same reference system.
The process generates a new aspect of the ancient map, showing the typical deformation
induced by its cartographic characteristics (and partly by the applied algorithm): in this way
it is possible to understand the metric quality of the map representation (e.g. by means of
the residuals errors associated to each single point, output by the geroreferencing process)
and the projection features of the historical map, but also to perform many other kinds of
analysis, e.g. studies related to change of the landscape.
In this specific case, after careful analysis of the three cartographic samples, a set of about 80
common GCPs, clearly identifiable also on the IGM (Italian Military Geographic Institute)
topographic sheet, was recognized on the three ancient maps. It has to be stressed that in
this phase a great deal of problems arose, concerning the basic characters of the points
themselves (e.g. their planimetric precision, their graphic representation on the ancient
maps, etc.), in addition to the difficulty in finding points that are still existing. North and
East coordinates were attributed to each selected point according to the UTM-ED50 (fuse 33)
grid. Different georeferencing methods were tested on the three map samples, and the most
useful in order to compare them with the present landscape resulted polynomial
transformations. In particular, the second order polynomial transformation resulted a good
compromise between adaptation of the ancient maps to inland area details of the present
landscape, on one hand, and constraint of deformations (indicated by the mean residual
errors) throughout the maps, on the other (Bitelli et al. 2009).
Polynomial transformations coincide with a linear transformation (6-parameter affine) in the
first order, and a non-linear one at higher degrees. A linear transformation corrects for scale,
offset, rotation and reflection effects, whereas a non-linear transformation (for example, the 2 nd
order polynomial transformation) corrects for non-linear distortions: the final result depends
very much on the number of control points and their spatial distribution in the image plane.
In Figure 3 an overlay of the three maps on present high resolution satellite images (in Bing
Maps TM environment) is reported. The mean residual error (expressed as RMS, Root Mean
Search WWH ::




Custom Search