441
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
used to fill the occluded areas. This is illustrated in Figure5.
Both points A, B have the gray value of pint B.
Perspective
Center
Figure5: Double mapping.
To restore the occluded areas, it is necessary to detect them and
use neighboring overlap images to extract the information of
these areas for replacing them in the orthophoto.
• Distance to occluded areas: smaller distance to
occluded areas will result in a disharmonious image.
• View angle: narrowest view angle will avoid
smearing problems.
When the images are mosaicked the differences between
radiometric characteristics of each image patch make seam lines
visible. Consequently, the histogram matching is inevitable to
equalize the radiometric responses of the image patches in order
to achieve a true orthophoto with high quality.
Perspective
Center
In the aerial Photogrammetry the relief displacements can be
decreased by using images created with a normal angle lens shot
from a higher altitude, but in the close rang photogrammetry
this condition cannot be easily fulfilled and hence large relief
displacements occur. So the occluded areas detection will be
time consuming and an efficient method must be used.
Figure 6: The principle of the height-based raytracing method
for occluded area detection.
4 EXPERIMENTAL RESULTS
3.1 Occluded areas detection
The detection of occluded areas is based on the visibility
analysis. If a ray from the output pixel intersects other features
in its path from DSM to perspective center, it is marked an
occluded pixel and filled with black color.
In this paper the height-based raytracing method is used for
detecting occluded areas (Bang, 2007). In this method, for each
pixel, the projection of ray onto the orthophoto plane (from the
nadir point to the pixel in the orthophoto space) is considered as
a search path. For testing the visibility of each pixel, the height
of the ray is compared with the height of some points along the
search path in specific intervals. If at least one point is found
with higher height than the ray height, the pixel will be marked
as occlusion. Figure 6 shows the principle of the height-based
raytracing method.
3.2 Mosaicking
In this stage the information of the occluded areas are restored
using the neighboring overlap images. In producing the true
orthophoto each occluded pixel in image have to be restored
with pixels in neighboring overlap images. So we must know
which pixel in which image must be used. This selection is
based on one or a combination of the following factors
(Schickler, 1998):
• Distance to nadir: smaller distance to nadir will result
in a smaller relief displacement for features not
modeled.
In this paper four close range images from four different
stations around the test object are used. The camera positions
are selected in such a way that by combining the images no
occluded areas remain. The positions of the stations are shown
in figure7.
Some retro-reflective targets are used on the test object to
determine the position and orientation of the camera. The DLT
equations are used as a camera model.