PHOTOGRAMMETRIC PROCESSING OF LOW ALTITUDE 
IMAGE SEQUENCES BY UNMANNED AIRSHIP 
Yongjun Zhang 
School of Remote Sensing and Information Engineering, Wuhan University, Wuhan, Hubei, 430079, P.R. China - 
zhangyj@whu.edu.cn 
Commission V 
KEY WORDS: Image Sequence, Close Range Photogrammetry, Matching, Aerial Triangulation, Accuracy Analysis, Topographic 
Mapping 
ABSTRACT: 
Low altitude aerial image sequences have the advantages of high overlap, multi viewing and very high ground resolution. These 
kinds of images can be used in various applications that need high precision or fine texture. This paper mainly focuses on the 
photogrammetric processing of low altitude image sequences acquired by unmanned airship, which automatically flies according to 
the predefined flight routes under the controlment of autopilot system. The overlap and relative rotation parameters between two 
adjacent images are estimated by matching two images as a whole and then precisely determined by pyramid based image matching 
and relative orientation. The matched image points and ground control points are then used for aerial triangulation with collinearity 
equations. The aerial triangulated high resolution images can be used to obtain precise spatial information products, such as Digital 
Surface Model (DSM), Digital Ortho Map (DOM) large scale Digital Linear Graphic (DLG) and three-dimensional (3D) city model. 
Experimental results show that the developed remote sensing system is qualified for high overlap and high resolution stereo imagery 
acquisition. Moreover, the proposed approaches are effective for photogrammetric processing of low altitude image sequences, and 
have well potentials in large scale topographic mapping and precise 3D reconstruction areas. 
1. INTRODUCTION 
In photogrammetric applications, image acquisition is one of the 
most expensive steps. Using a flexible and efficient way to 
obtain aerial images has been the primary purpose of many 
photogrammetric researchers and communities. 
As compared with film based camera, digital camera is 
advantageous for its ability of high forward overlap (e.g. 80%) 
image acquisition without increasing flight strips. In this case, 
every ground feature has at least 5 corresponding image features 
(Zhang, 2005a). High overlap image sequences are also 
advantageous for image matching and 3D reconstruction (Zhang, 
2005b). 
Along with the development of unmanned aerial vehicles, 
popularization of non-metric digital cameras and progress of 
photogrammetric technologies, low altitude images has been a 
hot spot in photogrammetric areas. Low altitude high resolution 
aerial images have well potentials in applications of large scale 
mapping, true orthophoto generation, archaeology and 3D city 
modeling. Now, the mostly used low altitude platforms for 
image data acquisition are helicopters (Zhang, 2005b), remote 
controlled model aircrafts (Nogami, 2002; Sun, 2003) or 
tethered balloons (Huang, 2003). 
Although considerable results of photogrammetric processing of 
low altitude images have been achieved (Karras, 1999; Bitelli, 
2004; Altan, 2004; Zhang, 2005b), the acquisition and 
photogrammetric processing of low altitude image sequences 
with unmanned platforms is still at an early stage. 
The typical characteristic of images taken from low altitude 
platforms is that the predefined relationship between adjacent 
images cannot be guaranteed strictly. There are usually large 
overlap variations and large rotation angles between adjacent 
images, and also large parallax discontinuities between features 
above the ground. So terrain continuity constraint that used by 
most traditional matching algorithms is invalid for low altitude 
images. 
General strategy for photogrammetric processing of low altitude 
image sequences, composition of the developed low altitude 
remote sensing system and data acquisition for experiments are 
described in section 2. Afterwards, detailed approach for 
photogrammetric processing, including image matching, 
relative orientation and aerial triangulation are discussed. 
Experimental results of photogrammetric processing and digital 
photogrammetric product generation are given in section 4. 
Finally, conclusions are given and further work is highlighted. 
2. GENERAL STRATEGY AND DATA ACQUISITION 
2.1 General strategy 
The photogrammetric processing of low altitude image 
sequences, such as image matching, relative orientation, aerial 
triangulation, DSM and DOM generation and large scale 
mapping are the content of the current work. The image 
sequences are automatically acquired by a low altitude remote 
sensing system based on unmanned airship. Firstly, the overlap 
and relative rotation parameters between two adjacent images 
are estimated by matching two images as a whole and then 
precisely determined by image matching and relative orientation 
process. Image matching is also constrained by epipolar line, 
which is determined by relative orientation parameters. The 
matched image points and ground control points are both used 
for aerial triangulation with collinearity equations. Orthoimage 
is rectified after DSM is generated by dense image matching