399 
STUDY ON AIRBORNE SAR IMAGE MATCHING USING EPIPOLAR GEOMETRY 
Xiaoman Luo 8, b , Guoman Huang 8 , Lei Pang 8 , Zheng Zhao 8 
a Chinese Academy of Surveying and Mapping, Beijing, 100039 
romanll21@hotmail.com, huang.guoman@163.com, panglei.mail@163.com, zhengzhaochina@163.com 
b Beijing SPOT Image Co. Ltd 
KEY WORDS: Airborne SAR, Image Matching, Radar Photogrammetry, Epipolar Geometry, Template Matching 
ABSTRACT: 
Aerial photos are broadly used to generate the 4D(DEM, DOM, DLG, DRG) products in photogrammetry, but there are some 
difficult areas where are usually rainy and cloudy, so in these areas optical image is hard to acquire. The airborne SAR has advantage 
to solve this problem, and in order to generate 4D products from SAR image, the study of Radar photogrammetry becomes more and 
more significant. This research is about airborne SAR image matching using similar corresponding epipolar line based on 
photogrammetry theory. First the epipolar geometry of aerial photo is analyzed, second SAR image is done some preprocessing (such 
as building pyramid-layered image and filter), third the image equation of stereo SAR is build to borrow optical method of epipolar 
geometry, finally result resampling image is test for image matching. This research is the basic preparation for building the stereo 
SAR, and makes the well technique foundation of generation 4D products. 
1. INTRODUCTIONS 
4D products are the basic digital product for national 
departments and various projects, which could offer lots of 
information. Optical image is usually used to generate the 4D 
products, but in some difficult areas, optical image is hard to 
acquire due to the rainy and cloudy weather. As an effective 
earth observation technique, Synthetic aperture radar (SAR) is 
being paid importance to and complementary to optical remote 
sensing owing to its all-weather and all- time capability. There 
are two important kinds of SAR platform: airborne SAR and 
space-borne SAR. The airborne SAR can acquire data 
according to requirement, and it’s more flexible than the 
space-borne SAR, which could choose the imaging target in 
pertinence. The resolution of the airborne SAR image is mostly 
higher than that of space-borne SAR. The more important 
characteristic is that it can offer the real time image for the 
disaster prevention and military reconnaissance. In order to 
generate 4D products from SAR image, the study of Radar 
photogrammetry becomes more and more significant. 
In the processing, the first step is to construct the stereo image 
pair. So using an appropriate method of image matching is 
important. 
Image matching is the process of transforming the different sets 
of data which were acquired by the same or different sensors, 
from the same or different viewpoint, from different 
perspectives geometry into one coordinate system. In a whole, 
like the traditional image registration methods, the means of 
SAR image matching can be classified into two categories: 
feature-based and intensity-based methods. Though the 
intensity-based image matching costs much time to compute, it 
is more reliable and it could keep the whole image information. 
In order to enhance the search efficient, the similar 
corresponding epipolar line method based on photogrammetry 
theory is used in this paper. 
the line where left or right photograph intersect with epipolar 
plane[l] (the plane including photographic baseline and ground 
point). According to the geometric definition of eqipolar line, 
the corresponding image points must in corresponding epipolar 
line. So it is more convenient to using corresponding epipolar 
line to search the corresponding image points. But the 
mechanism of the SAR imaging is different from the 
photogrammetric imaging. The appearance of SAR image in 
range direction is linear image, and in azimuth direction SAR 
image lays over the whole strip, so the whole SAR image is 
equal to multi-central projection linear array image which is 
similar to the Linear Array Push-broom image. According to 
the imaging equation of linear CCD Push-broom sensor, we can 
build the SAR row-center projection equation [2]. 
V 
Fig 1. Row-Center Projection 
x t 
*-x; 
° T 
1 
= xmJmJ 
Y-Y« 
_Z-Z sl _ 
2. THEORY OF EPIPOLAR GEOMETRY 
The eqipolar line is the basic concept of photogrammetry, it is