ANALYSIS AND THE SOLUTIONS FOR GENERATING A TRUE DIGITAL ORTHO
PHOTO IN CLOSE RANGE PHOTOGRAMMETRY
M. Shariat 3 *, A. Azizi b , M. Saadatseresht b
a Dept. of Geomatic Engineering, Faculty of Engineering, University of Tehran, Iran - mashariat@ut.ac.ir
b Centre of Excellence for Disaster Management, Dept, of Geomatic Engineering, College of Engineering, University of Tehran, Iran
-aazizi@ut. ac.ir
KEY WORDS: True Orthophotos, Occluded Areas, Grey Shade Interpolation, Histogram Matching, Mosaic, Close Range
Photogrammetry
ABSTRACT:
Digital orthophoto is an image where the relief displacements and the camera's tilt have been removed. So it is an efficient,
inexpensive and accurate approach for purposes such as evaluation, analysis or measurement of the objects presented in the images.
For this reason digital orthophoto has occupied an important place in spatial database in GIS. The traditional orthophoto is based on
Digital terrain model and it does not include the man-made objects. Therefore, the orthophotos of urban areas do not have
satisfactory accuracy. True orthophoto generation by using the digital surface model in rectification process and detecting occluded
areas through the visibility analysis is the only solution. True orthophoto can be produced from close range as well as aerial and
satellite images. In this paper, to simplify the true orthophoto generation process several close range images of a cubic object are
used with a regularly spaced grid DSM. In close range Photogrammetry the distance between camera and object is small and as a
result the occluded areas are extensive. Therefore, the occluded areas detection method must be efficient. In this research work the
height-based raytracing method is used to detect the occluded areas. To restore the information of the occluded areas the seamless
automatic mosaicking considering histogram equalization of images is applied.
1 INTRODUCTION
Today with increasing importance of GIS, particularly in urban
areas, the demands for digital orthophotos with high details and
greater accuracy are growing. Digital orthophotos provide
spatial and spectral information and have many applications
such as measurable virtual reality. There is not any relief
displacement in orthophoto and there are consistent scale in it
that cause to be similar to the maps. Since in the traditional
orthophoto production, DTM and aerial photographs of urban
areas are used and DTM does not contain man-made objects
such as buildings, bridges etc., and also because of the
perspective geometry inherent in the images, these features do
not locate in their true positions and hence a geometrical error
occur in orthophoto. This problem shows itself clearly if the
vector data of man-made objects is superimposed over the
corresponding orthophoto. Moreover, the man-made objects in
the image, produces occluded areas so that the quality of
orthophoto deteriorates. Therefore an orthophoto that is
geometrically complete and accurate is needed. This demand is
fulfilled with the so called true orthophoto that is produced
using the DSM. But when a DSM is used the double mappings
occur. These occluded areas must be filled using the
neighboring overlap images and a mosaicking procedure is
required.
In this paper, close range images of a cubic object are used.
Four images are taken from different direction by using a non
metric camera. One image is considered as a master image and
others are slave images. For recovering the position and
orientation of the camera during the exposure time some retro-
reflective targets as ground control points are used.
The final accuracy of the ortho-rectified image may be
improved by incorporating self calibration strategy. After the
computation of the exterior orientation and self calibration
parameters, the occluded areas are detected using the raytracing
approach. In the final stage, the image segments are stitched to
form a uniform image.
The sections that follow, after a brief review of classical method
of ortho-rectification, the above mentioned procedures for
generating true ortho are described and the final result is
presented.
2 TRADITIONAL ORTHOPHOTO
Two points with the same location and different height in the
object space will project at two different locations in image
space because of using the perspective projection. Differential
rectification is the process of removing these relief
displacements using the DTM, inner and outer orientation
parameters of image. The rectification can be done in two ways:
Forward projection and backward projection:
In forward projection for each pixel of master image, using
DLT equations, the ground coordinate is calculated. If the
ground coordinates of the comer points and the pixel size of the
orthophoto is known, the pixel coordinate in the orthophoto is
determined. In this way the three dimensional ground
coordinates of each image pixel is calculated using two
dimensional image coordinates and hence an iterative process is
required (Bang, 2007). Moreover, the regularly spaced points in
the image are projected to a set of irregular spaced points, so
they must be interpolated into regular points. To avoid aliasing
problem, the distribution of the regular points in object space
should satisfy the sampling theorem.
Figure 1 shows the basic geometry of the forward projection
approach for ortho rectification.