Building extraction from remote sensing imageries by data fusion techniques

IJRET: International Journal of Research in Engineering and Technology ISSN: 2319-1163
__________________________________________________________________________________________
Volume: 02 Issue: 03 | Mar-2013, Available @ http://www.ijret.org 347
BUILDING EXTRACTION FROM REMOTE SENSING IMAGERIES BY
DATA FUSION TECHNIQUES
Ezhili .G1
and Akshaya .V.S2
1.2
Computer Science and Engineering, R.M.D. Engineering College, Chennai, Tamil Nadu, India
ezhili1994@gmail.com, buvanaksh@hotmail.com
Abstract
This paper presents a data fusion approach for manmade objects extraction from high-resolution IKONOS satellite images. Buildings
can have various complex forms and roofs of various compositional materials. Their automatic extraction from imagery is a very
difficult problem. Applying normal image processing methods could not achieve satisfied performance, especially for high-resolution
satellite images. It is based on edge maps derived from IKONOS data. Local changes or variations of the intensity of the imagery
(such as edges and corners) are important information for image processing and pattern recognition. K-MEANS clustering is one of
the most popular techniques that can be used to classify satellite images. This technique coupled with canny edge detection, which has
double threshold technique is less fooled by noise, forms a very good tool in detection of man-made features. The above mentioned
techniques are applied to one meter IKONOS imagery of the highly urbanized Singapore city, to detect building edges within scene.
Keywords: Canny edge detection, RGB color matrices, Gaussian filter, K-means clustering, non-maximum suppression.
----------------------------------------------------------------------***------------------------------------------------------------------------
1. INTRODUCTION
Remote sensing is the science and art of obtaining information
about an object, area, or phenomenon through the analysis of
data acquired by a device that is not in contact with the object,
area, or phenomenon under investigation Cartographic
production facilities are involved in the creation of maps
containing features such as road vegetation boundaries, and
building footprints. The automation of such tasks can lead to
greater productivity, resulting in reduced timelines for map
production. This has significance for both military and civilian
emergency purposes. With the recent advent of a series of
commercialized high-resolution satellite, the potential of
IKONOS imagery in topographic mapping has been investigated
and highlighted by many researchers (Holland et al., 2002;
Holland & Marshall, 2003).However In urban areas, it has been
long a challenge topic to automatically extract urban objects
from images due to the high object density and scene
complexity. (Building).Buildings is the most relevant man made
structures. Their detection is valuable because of the strategic
human activity occurs in, are in association with, a building of
some sort. In addition as they do not move, they do not serve as
good references for the relative position of other type of objects.
But in a highly urbanized city, detecting individual buildings is a
problem, due to the proximity of the buildings to each other and
the heights of the buildings which causes relief displacement.
This paper deals with devolving an algorithm to rapidly and
automatically extract man made features (buildings) from remote
sensing imageries by data fusion techniques like k-means
clustering,feature extraction,edge detection, morphological
operations. The main difficulty of image segmentation is the lack
of adequate tools to characterize different scales of texture
effectively. Recent development in multi-resolution analysis
such as feature extraction helps overcome this difficulty. This
ability to discriminate features is generally dependent of scale.
Another difficulty with common edge operators is that they
detect too many images, which makes the map difficult to
interpret. The canny edge detector uses the threshold technique
by the noise is reduced and only the wanted edges are picked by
settling the thresholds.
2. METHODOLOGY
The description of the approach to discriminate man-made
objects from high resolution IKONOS data is shown in Figure.
Color to grayscale conversion
K-means clustering
Feature extraction
Edge detection
Morphological operations
Input image

__________________________________________________________________________________________
2.1 K means clustering technique
Image segmentation is the process of division of the image into
regions with similar attributes. K means clustering technique is
performed on the collection of remotely sensed data. In our trials
we have used 3 clusters. Centroids are initialized by finding the
mean vector and looking for those K vectors that are farthest
from the mean. Euclidean distance in the feature space is used as
the measure of dissimilarity. The convergence criterion is that
the difference in the centroids in successive iteration is less than
a predefined threshold. At the end of this run we get a class label
for each of the pixels and the centroids for each of the classes.
For the Kth cluster, the mean is given by,


kn
i
i
k
k x
n 1
1

Where µ k is the mean vector and n k is the number of vectors in
the K th cluster.
For the K th cluster, the covariance matrix is given by,
2
)
1
)(
1
k
n
i
i
k
k
k
x
n
C  
Where n k is the number of vectors in the K th cluster, x i is the
vector in the cluster k and µ k d is the mean vector of cluster K.
Mean and covariance values are refined in the Expectation
Maximization algorithm.
2.2 Texture Feature Extraction
Many land cover/land use classes in urban areas can be
distinguished from each other via their shape or structure
characteristics. Therefore, it is important to extract features that
are able to describe relevant "texture" properties of classes.
Texture is an important characteristic for the analysis of many
types of images such as an image obtained from aircraft or
satellite platforms. It is the visual effect, which is produced by
spatial distribution of tonal variations over relatively small areas
2 .The concept of texture can be investigated through its
relationship with spectral data in fact, textural and spectral
information can both be present in an image or either one can
dominate the other classification accuracy.
In the proposed algorithm for classification, the co-occurrence
features are selected as the basic texture feature detectors due to
their good performance in many pattern recognition applications
including remote sensing. A gray level co-occurrence matrix is
defined as a sample of the joint probability density of the gray
levels of two pixels separated by a given displacement. The
features based on GLCM are energy, entropy and correlation.
Gray-scale co-occurrence matrix Pd is obtained by following
computation.
The features computed are:
*
 
i j
d jiPEnergy ,
2
*
   
i j
dd jiPjiPEntropy ,log,2
*
    



i j yx
d jiPji
nCorrelatio

 ,
*
),()( 2
jiPjiInertia d
i j
 
*
 

i j
d
ji
jiP
ferenceInversedif 2
)(1
),(
*

i j
d jiijPationAutocorrel ),(
Where µ is the mean of Pd and σ y are the standard deviations of
Pd (x) and Pd (y) respectively.
3. THE CANNY EDGE DETECTION
Edges of an image reflect the information of the image mostly.
They contain the basis character of the image. Edges within an
image correspond to intensity discontinuities that result from
different surface reflectance of objects. Various illumination
conditions, or varying distance and orientations of objects from a
viewer. Edge detection is a common problem and of
fundamental importance in image analysis and computer vision.
Edges however generally occur at various resolutions, or scales,
and represent transition of different degrees, or gradient levels.
Perhaps the most commonly used method from detecting edges
in an image is through spatial gradient. In this approach, the
edges are identified by the local extrema in the differentiated
image through thresholding. The canny edge detector is less
likely to be “fooled” by noise and more likely to detect true
weak edges, which are very important for the detection of
building edges. The double thresholding of canny edge detector
plays the main role in edge detection. The method uses two
thresholds – to detect strong edges and weak edges, and includes
the weak edges in the output only if they are connected to strong
edges. The Canny operator works in a multi-stage process. The
Canny edge detection algorithm has the following steps:
        jvlIisrIvlsrPd  ,,,:,,,

__________________________________________________________________________________________
 Smoothens the image with a Gaussian filter.
 Computes the gradient magnitude and orientation using
finite-difference approximation for the partial
derivatives.
 Applies non-maxima suppression to the gradient
magnitude.
 Uses the double threshold algorithm to detect and link
edges.
The upper tracking threshold can be set to quite high and lower
threshold quite low for good results. Setting the lower threshold
too high will cause noisy edges to break up. Setting the upper
threshold too low increases the number of spurious and
undesirable edge fragments appearing in the output.
3.1 Morphological operations
The edge map obtained with above methodology is cleaned
using morphological operators to remove stray pixels and to
connect all those un-connected pixels. The various operations
include, cleaning, spurring, removing, bridging, etc. The
explanation for these operations is given below.
Bridge-bridge previously unconnected pixels
'Clean' - Remove isolated pixels (1's surrounded by 0's 'Close'
- Perform binary closure (dilation followed by
erosion)
'Diag' - Diagonal fill to eliminate 8-connectivity of
Background
'Dilate' -Perform dilation using the structuring element
ones (3)
'Erode' -Perform erosion using the structuring element
ones(3)
'Spur' -Remove end points of lines without removing
small objects completely.
Fig Caption 1: Depicts the original input image taken from
ikonos
Fig Caption 2: Depicts the clustered image of the input image
by k-means algorithm.
Fig Caption 3: Depicts the extraction of building features from
the clustered image
Fig Caption 4: Depicts the edge detected image of Fig Caption
3 image
Fig Caption 5: Obtained by morphological operation to the
Edge detected image from Fig Caption 4

__________________________________________________________________________________________
3.2 Results and discussion
Data characteristics
This one-meter resolution color image of the city of Singapore
was collected August 9, 2000 by Space Imaging's IKONOS
satellite. The entire image depicts the trade and administration
centers located between the mouth of the Singapore River and
Canning Hill. This IKONOS imagery is used for project
planning and monitoring, seaport and airport management,
insurance and risk management, disaster assessment, forestry
management and environmental monitoring, coastal zone
mapping, urban planning, and tropical vegetation studies.
CONCLUSIONS
A new technique for the extraction of buildings from satellite
images was presented. Promising results were obtained with our
preliminary experiments. Further testing of the algorithm is
currently under-way. It was shown that the algorithm could be
extended for the detection of land cover changes from remotely
sensed data. This can be used for project planning and
monitoring, seaport and airport management, insurance and risk
management, disaster assessment, forestry management and
environmental monitoring, coastal zone mapping, urban
planning, and tropical vegetation studies.
REFERENCES
[1] Roux & McKeown, 1994] M. Roux and D. McKeown,
“Feature Matching for Building Extraction from Multiple
Views”, Proceedings, IEEE Conference on Computer Vision
and Pattern Recognition, 1994, pp. 46-53.
[2] www.fp.ucalgary.ca/mhallbey/tutorial.htm
[3] Kiwon Lee, Byung-Doo Kwon “Urban Feature
Characterization using High-Resolution Satellite Imagery:
Texture Analysis Approach” Map Asia Conference 2004
GISdevelopment.net.
[4] N.Abbadeni, “Computational measures corresponding to
perceptual textural Features”, Proceedings of the IEEE
International conference on Image Processing,pp 897-
900,2000.
[5] R. L. Kettig and D. A. Landgrebe “Classification of
Multispectral Image Data by Extraction and Classification
of Homogeneous Objects” IEEETransactions on Geoscience
Electronics, Vol. GE-14, No. 1, pp. 19-26, January 1976
[6] R.Chellappa, R.L.Kashyap & B.S.Manjunath, “Model-
Based Texture Segmentation and Classification”, Chapter
2.2 in the Handbook of Pattern Recognition and Computer
Vision, C.H.Chen, L.F.Pau and P.S.P.Wang (eds.), World
Scientific Publishing Company, pp. 277-310, 1993.

Building extraction from remote sensing imageries by data fusion techniques

More Related Content

What's hot (16)

Viewers also liked (20)

Similar to Building extraction from remote sensing imageries by data fusion techniques (20)

More from eSAT Publishing House (20)

Recently uploaded (20)

Building extraction from remote sensing imageries by data fusion techniques