À 4000 to -100
À -100 to -50
À -50 to -20
-20 to 20
20 to 50
50 to 100
100 to 2000
ot size = 7 pixels;
the shadow zones,
lata. Such gaps are
ie grid are plotted,
ironment and, as;
reas on the object
1g is improved by
ide 50 cm, further
al vectorisation of
of the façade. À
that increasing the
cient to enter a set
ble value for this
ning at the end of
ous points with
a result, the DSM
points with high
d for each point is
:d by blank spaces
his. The matching
'se data.
| a high matching
; not an effective
; representation of
sd.
Bitelli, Gabrielle
12000
X (mm)
0.90 to 1.00
Figure 4 — Point values of the matching coefficient for the DSM generated with spot size = 21 pixels.
The object surfaces can be reconstructed by a set of points constituting point entities (DSM grid) and/or linear entities.
The latter allow the identification of object elements such as discontinuities or special geometric features.
In close-range applications, the a priori description of the object must be carefully developed, to avoid gaps in the data
which are reflected in the final orthophoto as surface deformations. During the plotting stage, ensure that all main
discontinuities are represented by inserting points, lines and polylines, having a clear advance idea of the type of
interpolator that will be used to create the DSM, support for the orthophoto.
In order to obtain a grid with user defined interval, an interpolating function determines the height value at the nodes on
the basis of the heights of the adjacent known points, assigning each point a weight which is proportional to the inverse
of the square of the distance from the node on the grid. The search can be limited to a surrounding area which is a
function of the curvature of the object surface, calculated on the basis of the starting points.
Otherwise, particularly if the available information includes lines and polylines, Delaunay triangulation is a model
widely used. This model is applied to the original data and leads directly to the construction of the object surface using
triangles. These triangles are generated in such a way that they have equal angles as far as possible and the smallest
possible sides, and the triangular Delaunay mesh passes through all of the object points.
Considering the two different approaches, and in order to identify the best procedure for creating an orthophoto of the
object surveyed, taking care to limit deformations on the image and the operator's work time, three methods for
orthophoto production have been tested on the study object.
Orthophoto I (figure 5aA): DSM in automatic mode according to a grid spacing 5 cm; search window size 7.2 cm on
object, correlation coefficient = 0.7; basic points 334. A full 5 cm DSM is obtained by interpolating values where not
previously automatically calculated and finally the model is resampled to a 50 cm mesh for orthophoto production.
Orthophoto II (figures 5aB, 5bA): the DSM is interpolated according to a grid spacing of 5 cm, taking in account vector
object plotting (points, lines and polylines) and resampled to 50 cm mesh for orthophoto production.
Orthophoto III (figure 5bB): the surface model is realised by a triangular mesh, directly from vector plotting data (lines,
polylines).
An examination of the results allows some conclusions to be drawn, although they do not apply in general, being linked
to the software platform used:
- the most effective procedure for creating an orthophoto of an architectural object, characterised by a number of planes
and by discontinuities, is to vectorise the main breaklines and create the surface which represents the object using
triangulation; projecting portions of the rectified image onto the surface generated in this way;
- a dense grid obtained from the interpolation of vector plotting does not always give an object surface without
deformations and, as a result, a metrically valid orthophoto. The problem arises from the interpolation used which, on
the basis of any algorithm, chamfers discontinuities such as boundary edges, indentations, etc. As a result, the surface
derived from the DSM thus obtained cannot be used to model the object correctly, generating visible deformations on
the orthophoto created;
- during the plotting stage, subsequent use of the numeric database must be clear, in order to adopt the most effective
plotting method (profiles, sections, grids, isolines, polylines). It is essential to know whether or not the data will be
interpolated and, if so, which model will be used. For example, if Delaunay triangulation is to be applied to the vector
plotting data, the edges of all surfaces must be identified, by entering lines and polylines. The only use of points can
create interference and must, therefore, be avoided. The lines and polylines must be accurately entered in order to
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000. 67
