International Journal of Computer Applications (0975 – 8887)
Volume 18– No.7, March 2011
10
Automatic Road Extraction based on Normalized Cuts
and Level Set Methods
M.Rajeswari
Department of
Telecommunication
Engineering
Bangalore Institute of
Technology
Bangalore560 004 KA, India
K.S.Gurumurthy
Department of Electronics &
Communication Engineering
University College of
Engineering
Bangalore 560 001 KA,India
L.Pratap Reddy
Department of Electronics &
Communication Engineering
JNTU College of Engineering-
Kukatpally,
Hyderabad-500 085 AP, India
S.N.Omkar
Senthilnath.J
Department of Aerospace
Engineering, Indian Institute
of Science
Bangalore, 560 012,
KA, India
ABSTRACT
Automatic road network extraction based on high resolution
satellite image for urban planning holds great potential for
significant reduction of database development/updating cost
and turnaround time. Satellite remote sensing has been
recognized worldwide as an effective technology for the
monitoring and mapping the urban development. Two
approaches for road network extraction for an urban region
have been proposed. When an image is considered in original
form it is difficult and computationally expensive to extract
roads due to presence of other road-like features with straight
edges. Hence roads are first extracted as elongated regions by
removing bright regions (that mostly represent the buildings,
parking lots and other open spaces), non-linear noise segments
are removed median filtering (based upon the fact that road
networks constitute large number of small linear
structures).The roads are then modeled as boundaries and are
extracted using Level set and Normalized cuts methods
.Finally The extracted roads are overlayed on the original
image. The experimental results show that these approaches
are efficient in extracting road segments in urban region from
high resolution satellite images. Evaluation of the results
carried out by comparing the level set and normalized cuts
results with manually extracted reference data. The methods
were applied on the high resolution IKONOS image of urban
area of Hobart, Australia.
Keywords
Level set, median filtering; Normalized cuts; Performance
Evaluation; Urban Road extraction;
.
1. INTRODUCTION
Road extraction from remotely sensed images has always
been a challenging problem. Fully automatic extraction of
roads from satellite imagery eliminates the need for human
operators to perform the time consuming and expensive
process of mapping roads from photographs. As increasing
volumes of high spatial resolution satellite imagery (e.g.,
Ikonos, QuickBird, OrbView-3, etc.) become available there
is a need for automation to extract information and analyze
image content.
Automatic detection and extraction of roads from remotely
sensed imagery has been an active research and development
topic for the last twenty years and in the practice for the last
twelve years, since the 1m resolution world’s first commercial
remote sensing satellite (Ikonos) become available in 1999.
Until now, various road extraction methods have been
proposed. For rural areas Zhang, C., 2004 [1] have done
database verification and updating determining the region of
interest for roads by a multispectral classification and
excluding high regions using Digital Surface Models then
parallel edges are extracted in the regions of interest. Mena,
J.B. and Malpica, J.A., 2005 [2] have used three different
classification methods for color and texture and are combined
to extract road regions. Roads are only extracted in the regions
around database roads. Road extraction is difficult in the
presence of context objects such as buildings or trees close to
the road, disrupting the appearance of the road or occluding it.
Gerke.M and Heipke.C 2008[3] and Baumgartner A.et al., [4]
have considered context objects. For urban areas Zhang Q and
Couloigner I 2006[5] have developed their extraction based
on multispectral classification and filtering using shape
criteria. Digital surface model (DSM) from LIDAR is used as
an additional data source by Hu.X et al., 2004 [6] to restrict
the search space for the straight lines. This cannot handle
curved roads well. In region based extraction Doucette, P et
al., 2001 [7] use hyper spectral data channels to extract road
regions and road pixels are grouped into a network with a k-
median classification. Hu, J et al., 2007 [8] extract road
footprints based on their shape and then track them. Junction
footprints are distinguished from ordinary road footprints.
They have used Post-processing to remove false extractions.
Bacher, U. and Mayer, H [9] calculates value for the road
class pixel then road hypotheses is determined using fuzzy
logic. Road network is generated using weighted graph and
detour factor to close small and large gaps respectively.
Poullis, C. and You, S., 2010 [10] have extracted road using
Gabor filter for image classification into road and non-road
pixels segmentation using graph cut and Gabor filter for post
processing Asef Zare [11]use feature extraction
(preprocessing), neuro-fuzzy system modeling and post
processing.
Recent methods for extraction of roads from high resolution
imagery include Normalized cut and Level set segmentation
methods. Level set method is a search algorithm that
determines evolving curve’s boundary pixels the level set
propagates as long as the speed function is greater than zero
.For the road extraction problem speed function has to be
greater than zero, at the edges of the true road boundary.
Therefore Level set is an efficient technique for extracting
road. Normalized Cuts is a graph-based method taking both
local and global characteristics of the image. Local
characteristics are incorporated into the similarity matrix for
considering the similarity of pixels in a close neighborhood.
Global characteristics are used for computing the best cut.
The combination of local and global aspects ignores noise,
small surface changes and weak edges and producing
extraction with most segments covering only a road area.
This makes normalized cuts suitable for automatic road
extraction.
In the literature on level set algorithm and normalized Cut for
road extraction. Rajeshwari, M et al., [12],Omkar et al.,[13]