523 
ROAD EXTRACTION FROM HIGH RESOLUTION REMOTE SENSING IMAGE BASED 
ON MATHEMATICS MORPHOLOGY AND SEED GROWTH 
Hongbin Ma Yahong Zhao, Qun He 
Surveying and Mapping Engineering Department of Northeastern University, 
Liaoning, Shenyang 110004, China.- neu_mhb@163.com 
Commission III, WG III/5 
KEY WORDS: road extraction, high resolution remote sensing image, seed growth, mathematics morphology 
ABSTRACT: Extracting the targets information from the remote sensing image has become the important method of updating the 
spatial geography information. With the development of spatial technology, sensor technology,digital image processing technology 
and the computer pattern recognition technology, how to extracting the targets information from the high resolution remote sensing 
image has become the important research content of the spatial information renews. And the exhaustive and accurate road net 
information plays an important role in the traffic control, the urban planning, the automatic vehicles navigation, the emergency 
business processes and so on. The Quick bird high resolution remote sensing image of Shenyang which is obtained in 
September,2006 is used as the research data. First, carry on pre-treatment to the remote sensing image, mainly uses the colour 
transformation-hsv transformation; Second use the supervised classification method- Support Vector Machine to classify, and 
evaluate the precision of the classification result, then change the classification image into binary image, and use the mathematics 
morphology method-(open\close operation)simplification image data, to maintain their basic shape characteristic, and except the 
irrelevant structure characteristic, in this paper we select the structure elements se0=strel('line', 10,30) and sel=strel('line',10,120) 
shape operation to the binary image, extract the road skeleton; at last use the seed growth algorithm to extraction the road median line 
that has certain length and direction, the experiment proves that This method that gives priority to mathematics morphology and 
gives assistance to the seed growing method has extract the road net information well, specially has the superiority in extracting the 
road detail information. 
1. INTRODUCT 
1.1 Perface 
As the important information source of digital photographic 
survey, feature extraction from the aerial image is the 
international development front topic of photographic survey, 
the remote sensing and the computer vision has the extremely 
important theory and the practical significance. With the 
development of spatial technology, the sensor technology and 
the computer technology, how to clear extract the targets 
information from the high resolution remote sensing image has 
become the important research content of the spatial 
information renewal. During the information time, how to 
automatic process, cognition and interpretation magnanimous 
image data is the important question during the entire social 
information process. The urban road is taken as the city skeleton, 
hand the pivotal status without doubt in the economic activity in 
the city. The high accuracy, the prompt renewal road net 
information plays an important role in the traffic control, the 
urban planning, the automatic vehicles navigation, the 
emergency business processes and so on [IJ .however, as we 
extract road information from high resolution remote sensing 
image ,the road width broadens and receives more serious 
disturbance, which increases the difficulty of extracting 
information. 
1.2 The research present situation 
Domestic and foreign has already many research of road 
extraction also has made very many achievements. And the 
correlation scholars have carried on the summary regarding 
this [2][3][4 l Jeong-Hun Jang et al [ 51 .first extract the center line of 
the straight-line band , then detect the different types of straight 
line band in this foundation , through the method of distance 
transformand so on ;Donald and Bruno [<l] have thoroughly 
discussed road recognition method of 10 meters resolution 
satellite image in the multitudinous research foundation; Gruen 
and Li [7] has used the GIS data to extract the linear features 
from the digital picture ;Zhang, C., Baltsavias, E [8j extract 
road network information from high resolution Aerial image 
based on Mathematical Morphology; Teger, Mayer and Radig [91 
extract road network using class and fuzzy class;Tupin,et 
al. [10] ,.extract road features from SAR images: using random 
field model; Katartzis,et al. [11] ,first use the soft mathematics 
morphology method to extract all road sections, then carry on 
liking the median line of the road section using the Markov 
Random Field Model , at last obtain the road net; Chanussot 12] 
detects linear object in SAR image using fuzzy fusion 
technology;Yuille and Coughlan [l3] analyzed and research the 
method of the automatic road features using probability 
methods,such as Bayesian,max-min-estimation ,the condition 
distributed and so on,through establishing structure tree; 
Laptev,et al. il4) research semi-automatic Linear Feature 
Extraction by dynamic Programming and LSB-Snakes from 
aerial image;Hu,et al. ll5] research semi-automated road 
extraction from aerial image based on Remote Sensing, 
template match and neural network;Shi Wenzhong,et 
al. ( 171 extract road network by the method that straight line 
matching combines with Mathematical Morphology post 
processing.But, at present, the researches of the road features 
extraction mainly aim at the aerial image and low, median 
resolution satellite remote sensing image (resolution is lower 
than 10 meters),the researches of the high resolution remote 
sensing image are few.