549 
AUTOMATIC ROAD EXTRACTION FROM HIGH RESOLUTION SATELLITE IMAGES 
USING NEURAL NETWORKS, TEXTURE ANALYSIS, FUZZY CLUSTERING AND 
GENETIC ALGORITHMS 
M. Mokhtarzade 3 ’ *, M. J. Valadan Zoej b , H. Ebadi b 
d Dept. of Geomatics Engineering, K.N. Toosi University, Tehran, Iran-m_mokhtarzade@yahoo.com 
b Dept. of Geomatics Engineering, K.N. Toosi University, Tehran, Iran-(ValadnZouj, Ebadi)@kntu.ac.ir 
Commission III 
KEY WORDS: Road Extraction, Neural Networks, Co-occurrence Texture Analysis, Fuzzy Clustering, Vectorization 
ABSTRACT: 
In this article, a new method for road extraction from high resolution Quick Bird and IKONOS pan-sharpened satellite images is 
presented. The proposed methodology consists of two separate stages of road detection and road vectorization. Neural networks are 
applied on high resolution IKONOS and Quick-Bird images for road detection. This paper has endeavoured to optimize neural 
networks’ functionality, using a variety of texture parameters. These texture parameters had different window sizes and grey level 
numbers, not only from source but also from pre-classified image. Road vectorization is based on the idea of road raster map 
clustering obtained from the previous road detection stage. In this step, despite of genetically guided clustering, a new flexible 
clustering methodology is proposed for road key point identification. The last step of road key point connecting is carried out based 
on the obtained information from a fuzzy shell based clustering. The accuracy assessment of the obtained vectorized road network 
proved the ability of the proposed method in sub-pixel road extraction. 
1. INTRODUCTION 
1.1 Road Extraction 
The presence of high resolution satellite images and their 
potential to be used in wide variety of applications such as 
preparing and updating maps have made the automatic 
extraction of object, especially roads and buildings, a new 
challenge in remote sensing. 
Traditionally, road extraction from aerial and satellite images 
has been performed manually by the operator. Considering the 
fact that this method was costly and time consuming the 
efficiency was by no means very high. 
Automatic road extraction provides means for creation, 
maintaining, and updating transportation network. It also 
provides data bases for traffic management, automated vehicle 
navigation and guidance. 
Vigorous methods have been proposed for automatic and semi 
automatic extraction of road networks from satellite images. 
Recently, these methods are more focused on high resolution 
satellite images due to their outstanding characteristics in 
mapping from space. 
1.2 Related Researches Review 
A comprehensive review on the proposed methods for road 
extraction is found in (Mena, 2003) where these methods are 
categorized from different aspects. A comprehensive reference 
list is also accessible. 
(Mohammadzadeh et al. 2006) proposed a new fuzzy 
segmentation method for road detection in high resolution 
satellite images with only a few number of road samples. 
Afterward by using an advanced mathematical morphological 
operator, road centrelines were extracted. 
A road detection strategy based on the neural network 
classifiers was introduced by (Mokhtarzade and Valadan, 2007) 
where a variety of input spectral parameters were tested on the 
functionality of the neural network for both road and 
background detection. 
The idea of geometrical and topological analysis of high 
resolution binary images for automatic vectorization of 
segmented road networks was presented in (Mena, 2006). 
Robust polynomial adjustment was used for geometrical 
analysis while mathematical morphological operators were 
applied in topological adjustment. 
Recently, many researchers have tested the idea of using 
contextual information for improving segmentation process of 
road regions. 
The research presented by (Mena and Malpica, 2005) is a good 
example for exploiting texture information in road extraction. 
In his paper, Mena and Malpica, performed a GIS updating 
using the pre-existing vectorial information and the RGB bands 
of high resolution satellite or aerial images. The binary 
segmentation performed in his research was based on Texture 
Progressive Analysis - the three level of texture statistical 
evaluation - being developed based on evidence theory 
framework. Finally, through skeleton extraction and 
Corresponding author