Key Words: orthoimage, DEM, hiding relation, visibility, context mapping
Abstract
Traditionally, orthophotos or orthoimages are made on the base of regular grid
DEM, which implies that the terrain surface is smooth. But as for close range photographs of
buildings, the hypothesis of smoothness is no longer right. And the hiding relations between
object surfaces exist here and there. This paper puts forward a method of producing accurate
orthoimage of buildings, which is based on TINs and context mapping. With this method,
hiding relation can be considered exactly and relative constrain conditions can be included.
An orthoimage software-WuTINOrtho, which will be a module of VirtuoZo, has been
developed
1. Introduction
Orthophotos are important products of
photogrammetry. The making of orthophoto
is a complex procedure. At the era of
analytical photogrammetry, people make
orthophotos with orthophoto projector such
as Avioplan OR-1. OR-1 adopts strip-scanning
method to produce orthophotos, in
which the zoom factor and rotating angle of
Dove are controlled according to the DEM
data. The principle of OR-1 determines the
DEM must be regular grids. However,
regular grids can not represent accurate
object surfaces and hiding relations between
object surfaces, which are very important
for making orthophotos of large-scale
buildings. In fact, the problem also exists
when making orthophotos of large-scale
urban areas. Just for that the scale of aerial
photographs is not too large, the
displacements of building roof can be
tolerated.
The development of digital photogrammetry
and the advantage of digital orthophoto
provide a possibility to overcome the
problem. The digital orthophoto is also
called virtual orthophoto or orthoimage.
When making orthoimages, point-wise
processing mode is better than the strip-wise
mode in analytical photogrammetry, for the
displacement and visibility of image points
can be considered point by point.
In close range photogrammetry of buildings,
the geometric relations between buildings
are very complicated. There exist not only
breaking in depth direction but also
complicated hiding relations. As shown in
Fig.2.there are three surfaces.the two
convex surface hide other surfaces. In order
to make high precision orthoimage, the
surfaces have to be measured exactly, and
the orthoimage should be generated with
context mapping method. The pixels should
be resampled on best photo. When
generating orthoimage, not only the
visibility in ortho direction but also the
visibility in projective direction of
photographs must be taken into account.
Generally speaking, the making procedure
of high precision orthoimage can be
described as Fig.1:
Fig.1
The making procedure of close range orthoimage
The First step is orientation of photos,
including interior orientation and exterior
orientation, providing the conditions of
reconstructing the ray bundles of
photographing. If only one photo or only
one stereo module is adopted, orientation is
not a big problem. But when making a big
orthoimage, in which more than one photos
or stereo models have to be adopted, the
precision of orientation can be a big
problem, because any small difference of
orientation can be visible in the overlapping
area in mosaic image.
Triangle is the simplest data structure,
which is also most flexible. Triangles or
TIN can represent object surfaces in any
precision. A TIN can represents a smoother
surface. And several TINS can represent
disconnected surfaces.
In order to consider the hiding relation
between surfaces the depth maps of
orthoimage and original images must be
generated. When generating orthoimage,
only surface point whose depth is not larger
than the depth of orthoimage depth map be
generated. On the other hand, because every
orthoimage point can be projected onto all
images, that on which image the destination
pixel value should be resampled is a key
problem.
In close range photograph of building, dead
angles are inevitable, which appear as blank
area on orthoimage. So orthoimage have to
be fixed with image processing software
such as Photoshop. With Photoshop, we can
also adjust the image level or even change
the color of the image.
