THE STUDY ON AUTOMATIC AND HIGH-PRECISION RECTIFICATION AND 
REGISTRATION OF MULTI-SOURCE REMOTE SENSING IMAGERY 
ZHANG Jixian*, LI Guosheng^, ZENG Yu ** 
‘Chinese Academy of Surveying and Mapping,Beijing, 100039, P.R.China, stecsm@public.bta.net.cn 
? Resource & Information College, Petroleum University, Dongying, 257062, P.R.China 
* Dept.of Geoinformation&Science, Shandong University of Science & Technology, Shandong Province, 271019 
3 
P.R.China 
KEY WORDS: Remote Sensing, Registration, Rectification, Multitemporal, Multiresolution, Multisensor 
ABSTRACT: 
The high precise method of automatically rectifying and registering the multi-source remote sensing imagery, which are from 
different sensors, with different resolutions and taken from different time, is presented in this paper. First of all, the image to be 
registered is roughly rectified by using the polynomial model. Then, the feature points, which are distributing evenly over the 
roughly rectified image are extracted automatically. Guided by the extracted feature points, the homologous control points to be 
used in automatic registration are obtained by using the pyramid-layered template matching technique. Finally, both the geo- 
referencing image and the image to be precisely registered are divided into a number of triangular regions by constructing the 
Triangulated Irregular Network (TIN) on them, and the high precise rectification and registration can be fulfilled triangular region 
by triangular region. 
1. INTRODUCTION 
The high-precision rectification and registration technique of 
multisensor, multiresolution and multitemporal remote sensing 
imagery is playing an important role in information fusion, 
change detection, map update and monitoring of environment 
and resources. Conventionally, two methods are used in image 
rectification and registration, which are the method using the 
polynomial model and the method using the satellite imaging 
model, such as the collinear transform. The latter is based on 
the relation between the image space and the ground space; It is 
the rigorous description of the geometric relationship of the 
imaging space; The accurate sensor position, sensor attitude and 
a Digital Elevation Model(DEM) are needed during the process. 
With the invention of the new kind sensors and the change of 
the imaging mode, it is getting more and more difficult to 
obtain the precise sensor position and sensor attitude 
information that vary continuously with time. At the same time, 
a DEM that can meet the needs is hard to acquire sometimes. 
That is why the image rectification and registration method, 
which is based on the polynomial transform, is simple and has 
been used widely. However, with the improvement of the 
spatial resolution of the remote sensing imagery and the 
extensive application of remote sensing technology, this method 
can not meet the increasing needs of high-precision rectification 
and registration. How to realize the high-precision rectification 
and registration of multi-source remote sensing imagery has 
been becoming the problem urgently to be resolved. On the 
basis of the entire image been processed by the polynomial 
transform first, an automatic and high-precision rectification 
and registration technique dealing with multi-source remote 
sensing imagery and based upon the Triangulated Irregular 
Network (TIN) division is developed in this paper. This 
technique is composed of the following five steps (Figure 1): 
(1) Apply the rough rectification to the uncorrected image by 
using the polynomial model. 
856 
(2) Automatically extract feature points distributed evenly 
over the roughly rectified image. 
(3) Guided by the extracted feature points, obtain the 
homologous control points to be used in automatic 
registration by using the pyramid-layered template 
matching technique. 
(4) Divide the geo-referencing image and the roughly rectified 
image into a number of triangular regions by constructing 
the Triangulated Irregular Network (TIN) over them. 
(5) Fulfill the high precise registration triangle by triangle. 
  
Geo-Referencing Image Uncorrected Image 
' v 
Entire Image Registration Using the Polynomial 
Model 
| 
Homologous Control Point Selection Using The 
Pyramid-layered Template Matching Technique 
TIN Construction 
i 
High Precise Registration Triangle by Triangle 
i 
High Precise Registered Image 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Fig.1 Workflow of automatic and high-precise rectification 
and registration of multi-source remote sensing imagery 
Ini 
re 
co 
ge 
ari 
an 
ex 
fu 
tré 
pc 
ca 
te 
re 
in 
tre 
tW 
au 
th 
an 
re 
de 
us 
au 
to 
ur 
sh 
th