UPGRADING FUNDAMENTAL GIS DATABASES FOR NAVIGATION FROM HIGH 
RESOLUTION SATELLITE IMAGERY 
MA Li^’ *, LI Jiatian c , CHEN Jun b 
a School of Remote Sensing and Information Engineering, Wuhan University, Wuhan 430079, China - lymar@126.com 
National Geomatics Center of China, Beijing 100048, China 
c Faculty of Land Resource Engineering, Kunming University of Science and Technology, Kunming 650093, China 
Youth Forum 
KEY WORDS: Road Extraction, Updating, Upgrading, Navigation, High Resolution 
ABSTRACT: Upgrading fundamental GIS databases for navigation use is an important work for providing location-based services 
In this paper, we present a process for extraction road networks in urban city from panchromatic IKONOS imagery, which is one of 
the steps in a framework for extracting the roads for the upgrading of the existing data of a fundamental topographic database. Our 
four-stage process is including image classification based on SVM, road orientation estimation based on edge direction histogram, 
directional filter for the classified road pixels and the intersection extraction of road networks. An initial result is shown in this paper. 
Our next research may include matching of the extracted road nodes with the vector data in fundamental databases. 
1. INTRODUCTION 
2.1 Analysis of Urban Road Networks 
Upgrading fundamental GIS databases for navigation use is an 
important work for providing location-based services (LBS). 
Many of the work are carried out mainly by field investigation 
at present, which is much cost and time consuming. High 
resolution satellite images make it possible to do some of the 
work partly automatically, namely the extraction of roads from 
high resolution imagery. 
In our reasearch, we aim at extracting road networks in urban 
city for navigation use. We think road networks in city as net 
works consist with road grids.Many road extraction methods are 
studied in recent years. (Mena, 2003) made a classification for 
the state of the art on road extraction for GIS update; 
(Quackenbush,2004) reviewed the techniques for extracting 
linear features from imagery, An overview of object extraction 
and revision by image analysis can be found in (Baltsavias, 
2004). Some of the work are special focusing on road junc 
tion/intersection/crossing) extraction (Price, 2000; Barsi, 2002; 
Gautama, 2004; Koutaki, 2004; Ravanbakhsh, 2007). 
In this paper, we present a process for extraction road networks 
from panchromatic IKONOS imagery, which is one of the steps 
in a framework for extracting the roads for the upgrading of the 
existing data of a fundamental topographic database. A short 
summary of the framework is given in section 2. Our four-stage 
process on road extraction is described in detail in Section 3. 
Section 4 contains some results of experiments. The last section 
gives a summary and draws some conclusions for the presented 
work. 
2. FRAMEWORK 
In our approach, we make an analysis for the road networks in 
urban city. Based on the analysis of road properties, we design 
a framework for extraction urban road network. 
We assume road networks in urban city as a grid network 
approximately. This means that the extracted road networks are 
not very irregular, and some small variations of geometry 
attributes are available. Figure 1 shows a typical imagery of the 
road networks in urban city. 
In our research, we think the urban road networks in imagery 
have some properties as follows: 
a) Road sections are interconnected. They meet at 
intersections, which are the main nodes of road networks. 
b) Road sections are intersected perpendicularly 
approximately and several roads are parallel in road 
networks. 
c) As a whole, road surfaces have similar spectral attribute 
in imagery. 
d) Most of the roads are straight; several curved road 
sections are connected with other straight road sections at 
intersections. 
e) Some other objects are highly related with road, such as 
vehicles, barriers, shadows, trees and buildings, etc. 
Figure 1. Typical road networks in urban city 
337