Sergey Zheltov
3. OBTAINING A MODEL OF AN UNDERL YING SURFACE
In the presented work the situation is supposed when detecting object can be present at relatively smooth curved
surface. Usual way of background surface model acquisition is least square method of fitting to the data set of points,
which are exactly known to be not belonging to the object under detection. To find such data different methods of
machine vision can be used.
Various methods of surface points obtaining based on machine vision algorithms are beyond the scope of this paper. As
a result of such methods application an irregular set of points is acquired. Different interpolation methods can be used
for obtaining a regular grid of surface model. The well-known inverse distance weighting (5) can produce reliable
surface model.
X 7 — 4’ . : à. . 2 e T m. 2 y —} à. ^j _ 1 5
Z(X.Y) SZ wi Rj =( 2) 4( ;) M pis (5)
Here (X;, Y;, Z;) is a set of known points; Z(X, Y) is a surface model.
Another model, which can be effectively used especially in road scenes, is [S.M.Smith, 1995]
Z(X,Y) 2 ao * ajX * a;Y * a;Y,, (6)
where coefficients a; can be found by fitting the model to the known set of points using least squares method.
As an example we show a road model that comes from road signs and markings detection (see Fig.2).
"s
y
s
(b) (c)
M TIER (d)
Figure 2. (a) Test image (left image of the stereopair); (b),(c) result of road marking detection for left and right images
of the stereopair; (d) road model obtained from the result of road marking dete ction by inverse distance weighting.
4. ORTHOPHOTO DIFFERENCE GENERATION
Orthophoto is orthogonal projection of 3D scene in which all distortions caused by camera orientation and ones caused
by 3D-shape of scene object are eliminated. Once orientation parameters of the cameras and reconstructed surface are
known it is possible to generate orthophotos of the scene using left and right images (Fig.3(a,b)). The orthophotos are
generated in the interest area X,S X € X» Y, S Y S Y; of the base plane. Denote orthophotos generated using left and
1044 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.
ri;
11