Jeffrey Shan 
  
AUTOMATIC IMAGE ORIENTATION BY USING GIS DATA 
Jeffrey J. SHAN 
Geomatics Engineering, School of Civil Engineering 
Purdue University 
IN 47907-1284, West Lafayette, U.S.A. 
jshan @ecn.purdue.edu 
Working Group IC-11 
KEY WORDS: Sensor, Orientation, Database, GIS 
ABSTRACT 
This article addresses a methodology and the experience gained on automatic exterior image orientation by using 
current information existing in GIS database. The proposed approach automatically extracts and matches feature points 
in evenly distributed patches on aerial and ortho images. The elevation of successfully matched points on the aerial 
image will then be obtained through interpolation calculation on the DTM data in the database. Robust bundle 
adjustment with great redundancy is therefore conducted to estimate the exterior orientation parameters. An innovative 
algorithm is proposed for mono-image intersection by using thus obtained exterior orientation parameters and current 
DTM data. The proposed approach is evaluated based on manually measured a great number of check points. Tests and 
analyses showed that for the ground points obtained through mono-image intersection, the planimetric accuracy is better 
than Ipixel, which fulfills the requirement of orthoimage production for database updating purpose. The elevation 
accuracy is solely dependent on the current DTM and its interpolation applied on it, which reaches 0.00396 flying height 
in this relative flat test area. 
1 INTRODUCTION 
Research from national and private organization shows that orthoimage is becoming an indispensable component in GIS 
(Geographic Information System) and its updating cycle is shortening to 5- 4 years or less [3,5]. Therefore, efficient 
method is expected, which can best utilize existing database information, such as orthoimage and DTM (Digital Terrain 
Model) to update current database. Exterior image orientation is the first step in this process. 
One of the main concerns in an automatic process is its accuracy [2]. In order to reach high accuracy, an automatic 
approach is proposed in this article, which extracts and matches feature points from evenly distributed patches on the 
aerial and ortho images. In this way, thousands of image points are obtained for the subsequent bundle adjustment, 
where robust estimation is therefore introduced to further detect and eliminate mismatched points to ensure the quality 
of space resection for the determination of image exterior orientation. Evaluation on the proposed approach is 
performed on coordinates of ground points determined by using a novel mono-image intersection algorithm. In this way 
it provides a direct accuracy estimation for the updated orthoimage. 
The remainder of this article consists of three sections. Section 2 describes the methodology of the proposed approach 
for automatic orientation calculation. In order to evaluate the accuracy of image orientation and its influence on ortho- 
rectification for orthoimage production in GIS database updating, an innovative and accurate algorithm for conducting 
mono-image intersection by using DTM is proposed in this section as well. By using the test data provided by OEEPE, 
Section 3 presents the test results for feature extraction, correspondence, and matching; orientation parameters and their 
precision estimation. Evaluation on the efficiency and quality of the proposed approach is conducted in Section 4, 
where coordinates of ground points determined from mono-image intersection based on manual and automatic 
measurements are compared and analyzed. The article is summarized in Section 5 with concluding remarks. 
2 METHODOLOGY 
The proposed approach is based on image matching between aerial image and its corresponding orthoimage in GIS 
database. Successful matching with orthoimage will provide the image points on the aerial image with their planimetric 
ground location. Their elevations can then be obtained by subsequent DTM interpolation operations. Space resection is 
therefore conducted by including all successfully matched image points and their three-dimensional coordinates in a 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 831