ROAD EXTRACTION FROM HIGH-RESOLUTION 
REMOTE SENSING IMAGE BASED ON PHASE CLASSIFICATION 
Cui Ni a *, Qin Ye a , Bofeng Li a , Shaoming Zhang 3 
3 Dept, of Surveying and Geo-informatics Engineering, Tongji University, 1239# Siping Rd., Shanghai 200092, China 
-nini_tong@ 163 .com 
Commission HI 
KEY WORDS: Road network extraction, High-resolution remote sensing image, Characteristic extraction, Phase-based 
classification, Gray level morphology, Binary image 
ABSTRACT: 
It is still an open problem to extract road feature from high-resolution remote sensing image, although this topic had been intensively 
investigated and many methods had been put forwards. All works for this thesis are focused on modem urban road and include the 
following four steps: image pre-processing, threshold calculation, feature extraction for straight line and curved line, target 
reconstruction. In this contribution, a new and semi-automatic approach is proposed based on phase classification. Firstly, basic road 
network can be obtained from high-resolution remote sensing image based on grey level mathematical morphology and canny 
algorithm and then road information can be exactly extracted by means of the “grey” parameters which are various for different 
kinds of road models based on the theory of phase-based classification. Additionally, the proposed method can also be employed to 
elevate urban highways, especially for the curve parts of which. The extracting results are reasonable. 
1. INTRODUCTION 
The high-resolution remote sensing image is usually refers to 
the image with its spatial resolution of the pixel below 10m. In 
the 1960s, the research on special high-resolution sensor 
reached a plateau of development and a lot of sensors with 
high-resolution were invented at that time, such as IKONOS, 
QUICKBIRD and so on. With the rise of special resolution, 
more and more details will be showed clearly, which could 
provide a reliable foundation for the work of extracting objects 
from high-resolution remote sensing image accurately. 
With the development of city, the objects become various, like 
building, park, athletic field, open area and so on and each one 
of which has itself complex structure and shape. It is rather hard 
to extract the different objects one by one and just do cost much 
time, and therefore, the extracting work can barely be set up in 
application. However, we all know that urban road is the main 
clue to analyze and interpret the city. So we can extract road 
information from urban remote sensing image first, and then 
analyze each district segmented by roads. By doing this, the 
analyzing work will be easier than before. 
There have been many researches on the approaches of road 
extraction (Shi et al. 2001). These can be divided into the semi 
automatic type (Li et al. 2004) and totally automatic type. 
Semi-automatic extraction for road feature takes advantage of 
man-machine interactive form to recognize the objects. Its main 
idea is consist of two steps. At first, give the initial pixel 
manually, sometimes even the initial direction. Finally process 
the image by computer. Studies on the aspect have received 
accurate effect. We can arrange it under the following 
approaches roughly. That is, road extraction from remote 
sensing image based on two-dimension wavelet transform(Zhu 
et al. 2002); Road extraction based on pixel and background 
arithmetic operator; Road extraction based on tree-structure 
model of feature cognition, which is fit for middle & low 
resolution image; Road extraction based on least square B- 
spline curve; Road extraction based on heuristic graph search . 
This approach has strong noise immunity; Road net extraction 
based on class and fuzzy sets (Gruen et al. 1995); LSB-Snakes 
approach (Trinder et al. 1997). 
At present, there isn’t a more generic road extraction system 
with complete autoimmunization. But some successes on a 
particular type of road extraction have been achieved, from 
which many meaningful algorithms have been attained. For 
example: road extraction based on parallel lines; road extraction 
based on the bi-value knowledge(Song et al. 2005); road 
extraction based on the characteristic of the window model; 
Besides described above, there are also other algorithms on 
automatic feature extraction. In the thesis” Fundamental Limits 
of Bayesian Inference: Order Parameters and Phase Transitions 
for Road Tracking” (Yuille et al. 2000), a popular approach has 
been studied. Because there is no exact image for experiment, it 
has more theoretical signification. 
Traditional approaches on road extraction tend to extract the 
linear road. The accuracy is relatively low in extracting the 
urban road. There also exist omissions in some images because 
of the tiny, closed and preserved features. Sometimes the road 
in the image is close to the background, which could go against 
extraction. And a suitable algorithm for urban highways in 
modem days hasn’t been applied in traditional approaches. 
Aiming to the shortfalls above, this paper proposes a new 
approach. That is road extraction based on Phase Classification. 
Not only be fit for the road extraction of general shapes, but 
also having certain superiority in the detection work of urban 
* Cui Ni, (1982-) Major in Photogrammetry & Remote Sensing, Tongji University, 1239# Siping Rd., Shanghai, China. Email: 
nini_tong@163.com. Tel. 13917330461 
509