In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
457 
ROAD EXTRACTION FROM ALOS IMAGES USING MATHEMATICAL 
MORPHOLOGY 
F. S. P. de Castro a ’ b *, J. A. S. Centeno b 
a IBGE, Brazilian Institute of Geography and Statistics, Porto Alegre R.G. do Sul, Brazil, - fabiana.piresc@gmail.com 
b UFPR, Department of Geomatic, Curitiba Paraná, Brazil - (centeno)@ufpr.br 
Commission VII 
KEY WORDS: Extraction, Identification, Programming, Image, Satellite, Mathematics, Method 
ABSTRACT: 
Over the past few years have seen the need to use remote sensing data to accomplish the complex task of automatic extraction of 
features. Among the sensor systems currently used for mapping can be highlighted the recent launches of new orbital satellites, for 
example, the Advanced Land Observing Satellite (ALOS). The problems currently involved in the extraction of features like road 
presents the following issues: The roads may be partially hidden and stretches of road may not be recorded due to limitations of the 
sensors. Given the need for analysis of the potential of ALOS images, development of methodologies for roads extraction, and the 
study of problems involved in the process, the aim of this paper is the roads extraction with ALOS images through the use of 
mathematical morphology. At first step, an initial selection of stretches of road is done using algorithms of mathematical morphology 
and segmentation. At this stage, most of other classes, such as vegetation were eliminated. However, at this moment the road had not 
complete obtained, performing inconsistently. This happens due to the spectral similarity between some sections of the road and 
vegetation present in the scene. Thus, to segment the image in order to eliminate the vegetation, parts of the road were also 
eliminated. In a second step, within the MATLAB environment, were developed a routine to complement the road obtained after the 
application of morphological operators. This routine used other techniques of mathematical morphology, and the Euclidean distance 
for complementation. At the end of the process, the road was complete, resulting in a road consistently. Further tests must still be 
performed, since the methods and techniques used to extract features modify by the area of study and the type of image being used. 
1. INTRODUCTION 
Given the continental dimensions of Brazil, a trend to research 
for a solution to the problem of outdated mapping in the 
country is the use of remote sensing data to accomplish the task 
of automatic extraction of features. According to Dal Poz et al. 
(2007), the problem of features extraction have been of 
fundamental importance for more than two decades at the 
automation of processes that extract cartographic features such 
as buildings, rivers, roads, etc. However, the developed 
solutions always depend on the type of sensor used to generate 
the images. 
New images are being made available with the recent launches 
of new orbital satellites, such as the Advanced Land Observing 
Satellite (ALOS). According to the Brazilian Institute of 
Geography and Statistics (IBGE, 2007), the ALOS images are 
intended mainly to serve the scientific community and the non 
commercial User, thus practicing a policy of cost where the 
images will have an affordable price. This cost policy is to make 
possible the mapping of a large country like Brazil. 
The roads are features on maps which can be highlighted for its 
dynamism due to changes in their shape or texture, type - 
pavement or not pavement - and / or inclusion of new roads or 
road sections in the system. These changes are constant due the 
result of transformation resulting from the socio-economic 
growth. Soon, the roads are cartographic features that require 
constant updating. 
The problems currently involved in the features extraction like 
road presents the following issues: first, the roads may be 
partially hidden; second, some stretches of road may not be 
registered due to limitations of the orbital images and third, the 
radiometric resolution of the selected road. The first issue is due 
to clouds that may be present in the images and shadows of 
structures such as buildings, bridges and cars as well as 
vegetation can hide parts of the road feature. The second 
question is a function of spatial resolution of the orbital images 
used for extraction. The third question refers to the radiometric 
similarity between different features, for example roads without 
pavement may present a feature similar to exposed soil. 
According Cleynenbreugel et al. (1990), one of the problems of 
roads extraction from satellite images is the spatial resolution of 
the image. This may involve many details of the roads were not 
visible in the images, so cannot be used for extraction. 
Solutions to the problem of roads extraction have been studying 
in different ways. The way to approach the issue has been 
modifying according to new sensors developing. Nowadays 
high spatial resolution images and laser scanner data are used. 
The difference between the proposals for roads extraction due to 
the strategy used, for example: type and resolution of the images 
are being used, configuration of the experiments, methods of 
preparation and general assumptions (Wang et al., 2005). 
Some solutions for roads extraction are described in the 
literature, as in Baumgartner et al. (1999) that used the same 
aerial image with different resolutions for an automatic road 
extraction based on multi-scale, grouping, and context. 
Wiedemann and Wessel (2003) extract roads from synthetic 
* Corresponding author. This is useful to know for communication with the appropriate person in cases with more than one author.