. r' ' . 
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing 2008 
As can be seen from the experimental results of aerial 
triangulation, when compared with the 0.05m GSD, height 
precision of about 0.4 and 0.6 GSD can be achieved by 75% and 
50% side overlap non-metric image sequences respectively. 
Bundle adjustment with additional parameters can significantly 
improve the height precision. The achieved height precision is 
significantly higher than traditional aerial triangulation, and 
consists with the theoretical analysis in section 3.4. 
4.3 Digital products generation 
Once aerial triangulation is finished, DSM can be generated by 
dense image matching and forward intersection. Firstly, feature 
points are extracted every 11 by 11 pixels in each image. Then 
conjugate image points are found by multi-image matching 
algorithms under constraints of the matched image points for 
aerial triangulation and known camera parameters. Finally, 
corresponding ground coordinates of all conjugate points are 
obtained by multi-image forward intersection (Zhang, 2005a). 
These ground points can be used to generate Delaunay triangles, 
which is the common data structure to represent 3D terrain 
information. 
DOM can be obtained by ortho rectification from image 
sequences with known camera parameters and the above 
obtained DSM. Note that mosaic lines of ortho image should 
not go through objects above the ground such as buildings. 
Color and illumination are inevitably different among images 
especially images of different strips acquired by digital cameras. 
So the generated ortho image has to be further processed by 
dodging or other color adjustment techniques. Figure 9 shows 
the generated ortho image after color adjustment. The precision 
of DOM can be checked by known ground features such as 
GCPs. Coordinates of GCPs can be interactively measured from 
ortho image and then compared with the world coordinates 
measured by total station. Precision of the generated ortho 
image is about 0.045m, i.e. comparable with GSD, as compared 
with the world coordinates. 
Figure 9. DOM of the test area generated by image sequences 
DLG is another type of digital photogrammetric productions. 
The aerial triangulated image sequences are also used to 
interactively produce DLG with commercial Digital 
Photogrammetric Workstation (DPW) VirtuoZo. Figure 10 
shows the stereoscopically measured outlines of buildings from 
all images in the test area. Coordinates of GCPs are also 
measured by stereo environment and then compared with world 
coordinates. Both planar and height precision are better than 
0.05m, which means that the proposed approaches and the 
image sequences acquired by the low altitude remote sensing 
system are qualified for the precision requirements of 1:500 
scale mapping. 
5. CONCLUSIONS 
The principles and technologies of photogrammetric processing 
of low altitude aerial images acquired by unmanned airship are 
discussed in this paper. The advantage of low altitude image 
sequences is that forward and side overlaps are both higher than 
that of traditional photogrammetry. Furthermore, geometric 
model of low altitude image sequences is also stronger. So the 
precision and reliability of aerial triangulation are higher since 
there are more redundant observations. 
The proposed image matching approach can automatically find 
conjugate points from image pairs with different overlaps and 
large rotation angles. High order correction polynomials used in 
aerial triangulation can significantly improve the height 
precision. First results of image matching and aerial 
triangulation are very satisfying. Precision of ortho image and 
DLG also qualifies for the criteria of 1:500 scale mapping. 
Experimental results show that the proposed approaches and the 
developed low altitude remote sensing system are promising in 
large scale digital photogrammetry and precise 3D 
reconstruction areas. 
Generating 3D models of buildings and other objects with low 
altitude image sequences will be the work in the near future. 
Introducing Global Positioning System (GPS) to measure 
coordinates of exposure points and thus decrease the number of 
required GCPs also need to be performed. 
ACKNOWLEDGEMENT 
This work is supported by National Natural Science Foundation 
of China with project number 406711-57 and 40620130438, 
Program for New Century Excellent Talents in University with 
project number NCET-07-0645, Hi-Tech Research and 
Development Program of China with project number 
2006AA12Z136 and National Key Technology Research and 
Development Program with project number 2006BAJ09B01. 
756