Extraction of Satellite Image using Particle Swarm Optimization

Abstract

Of all tasks in photogrammetry the extraction of cartographic features is the most timeconsuming. Fully automatic acquisition of features like roads and buildings, however, appears tobe very difficult. The extraction of cartographic features form digital satellite imagery requiresinterpretation of this imagery. The knowledge one needs about the topographic objects and theirappearances in satellite images in order to recognize these objects and extract the relevantobject outlines is difficult to model and to implement in computer algorithms. This paperintroduces Particle Swarm Optimization based method of object extraction from Google Earthimage (satellite image). This paper deals with the land cover mapping by using swarm computingtechniques. The motivation of this paper is to explore the improved swarm computing algorithmsfor the satellite image object extraction.

Full text

Page 1

Er. Harish Kundra, Dr. V.K. Panchal, Karandeep Singh, Himashu Kaura & Sagar Arora
International Journal of Engineering (IJE), Volume (4): Issue (1) 86
Extraction of Satellite Image using Particle Swarm Optimization


Er.Harish Kundra
Assistant Professor & Head hodcseit@rayatbahra.com
Rayat Institute of Engineering & IT , Railmajra,
Punjab,India.

Dr. V.K.Panchal
Director, vkpans@hotmail.com
DTRL,DRDO,
Delhi,India.

Sagar Arora
B.Tech Student touch_sagar@yahoo.com
Rayat Institute of Engineering & IT
Punjab,India.

Karandeep Singh
B.Tech Student karanatter@gmail.com
Rayat Institute of Engineering & IT, Railmajra
Punjab,India


Himashu Kaura
B.Tech Student himanshu.kaura21@yahoo.com
Rayat Institute of Engineering & IT, Railmajra
Punjab,India.


Abstract

Of all tasks in photogrammetry the extraction of cartographic features is the most time
consuming. Fully automatic acquisition of features like roads and buildings, however, appears to
be very difficult. The extraction of cartographic features form digital satellite imagery requires
interpretation of this imagery. The knowledge one needs about the topographic objects and their
appearances in satellite images in order to recognize these objects and extract the relevant
object outlines is difficult to model and to implement in computer algorithms. This paper
introduces Particle Swarm Optimization based method of object extraction from Google Earth
image (satellite image). This paper deals with the land cover mapping by using swarm computing
techniques. The motivation of this paper is to explore the improved swarm computing algorithms
for the satellite image object extraction.

KEYWORDS - Objects extraction, Google Earth image, Unsupervised Learning PSO algorithm



1. INTRODUCTION

Natural or man-made objects extracted from Google earth has been used for many different
purpose, e.g.. military, map publishing, transportation, car navigation, etc. Automatic natural
objects (road, building, forest, water body etc.) extractions are a challenging problem, and no

Page 2

Er. Harish Kundra, Dr. V.K. Panchal, Karandeep Singh, Himashu Kaura & Sagar Arora
International Journal of Engineering (IJE), Volume (4): Issue (1) 87
existing software is able to perform the task reliably. Since manual extraction of natural objects
from imagery is very time consuming, automatic method have the potential to improve the speed
and utility for military and civil application and are therefore highly desirable. Google Earth is a
virtual globe, map and geographic information program that were originally called Earth Viewer.
Google Earth provides satellite images of any area. In this paper, Google earth’s image is used
for implementation of extraction natural or man-made object from satellite image.Simultaneously
there is a new wide range of computational algorithms that have emerged from the behaviour of
social insects. Social insects are usually characterized by their self organization and with the
minimum communication or the absence of it. Every social insect individually is self-autonomous.
They can obtain information about environment and interact with the remote insects or
environment indirectly, by stigmergy. All these features characterize Swarm Intelligence . We can
found these features in nature such as ant colonies, bird flocking, animal herding, fish schooling
etc. The two most widely used swarm intelligence algorithms are Ant Colony Optimization (ACO)
and Particle Swarm Optimization (PSO).
We are presenting here the usage of PSO (Particle Swarm Optimization) algorithms applied to
Google earth’s image for the extraction of natural or man-made object. The activities for the
development of this paper are organized in four sections.
In section 1, a high- resolution satellite image (Google earth), including scenes of highways,
forest, buildings, were formed and how particle swarm optimization, unsupervised learning
technique are used. In section 2, using the collected image form section 1, PSO algorithm to
detect and extract the natural or man-made objects was developed.In section 3, algorithms to
detect, extract and classify natural or man-made objects (road, building, forest, barren land etc.)
were implemented.In section 4, the feasibility of using this image and the results are presented.

2. SECTION 1: USAGE OF IMAGE FROM GOOGLE EARTH AND
APPLICATION OF PARTICLE SWARM OPTIMIZATION AND UNSUPERVISED
LEARNING

Google Earth
Google Earth is used to get number of satellite images. Google Earth is a virtual globe, map and
geographic information program that was originally called Earth Viewer, and was created by
Keyhole, Inc, a company acquired by Google. To collect image from Google Earth and
implementing PSO algorithm for extracted natural or man-made object from satellite image
(Google Earth). Google Earth displays satellite images of varying resolution of the Earth's
surface, allowing users to visually see things like cities and houses etc. The degree of resolution
available is based somewhat on the points of interest and popularity, but most land (except for
some islands) is covered in at least 15 meters of resolution.
A few Google Earth images :
A) B)

Page 3

Er. Harish Kundra, Dr. V.K. Panchal, Karandeep Singh, Himashu Kaura & Sagar Arora
International Journal of Engineering (IJE), Volume (4): Issue (1) 88
Figure 1. Enlarge Google earth images (A,B)

Particle Swarm Optimization
Particle swarm optimization (PSO) is a swarm intelligence based algorithm to find a solution to an
optimization problem in a search space, or model and predict social behavior in the presence of
objectives.
Particle swarm optimization is a stochastic, population-based computer algorithm for problem
solving. It is a kind of swarm intelligence that is based on social-psychological principles and
provides insights into social behavior, as well as contributing to engineering applications. The
particle swarm optimization algorithm was first described in 1995 by James Kennedy and Russell
C. Eberhart. The techniques have evolved greatly since then, and the original version of the
algorithm is barely recognizable in current ones.
Social influence and social learning enable a person to maintain cognitive consistency. People
solve problems by talking with other people about them, and as they interact their beliefs,
attitudes, and behaviors change; the changes could typically be depicted as the individuals
moving toward one another in a socio-cognitive space.
The particle swarm simulates this kind of social optimization. A problem is given, and some way
to evaluate a proposed solution to it exists in the form of a fitness function (4).
The swarm is typically modeled by particles in multidimensional space that have a position and a
velocity. These particles fly through hyperspace and have two essential reasoning capabilities:
their memory of their own best position and knowledge of the global or their neighborhood’s best.
So a particle has the following information to make a suitable change in its position and velocity:
a. A global best that is known to all and immediately updated when a new best position is
found by any particle in the swarm.
b. Neighborhood best that the particle obtains by communicating with a subset of the swarm.
c. The local best, which is the best solution that the particle has seen.

Unsupervised Learning
In unsupervised learning, all the observations are assumed to be caused by latent variables, that
is, the observations are assumed to be at the end of the causal chain. In practice, models for
supervised learning often leave the probability for inputs undefined. This model is not needed as
long as the inputs are available, but if some of the input values are missing, it is not possible to
infer anything about the outputs. If the inputs are also modeled, then missing inputs cause no
problem since they can be considered latent variables as in unsupervised learning.
In unsupervised learning, the learning can proceed hierarchically from the observations into ever
more abstract levels of representation. Each additional hierarchy needs to learn only one step
and therefore the learning time increases (approximately) linearly in the number of levels in the
model hierarchy.

3. SECTION 2: DEVELOPMENT OF PSO ALGORITHM TO DETECT AND
EXTRACT NATURAL AND MAN MADE OBJECTS

Natural or man-made object Extraction.
Object extraction of image procedures is to automatically categorize all pixels in an image into
roads, forest, barren land, water bodies and settlements. In this paper we extract the objects from
the images using PSO algorithm. To do this, appropriate threshold values have to be computed,
and then segmentation is performed. To compute the appropriate threshold values several
problems had to be analyzed. In this phase, an algorithm for the extraction of objects from Google
earth’s image was developed. This test image is presented various conditions under which the
performance of our algorithm was evaluated. In the following section the proposed algorithm is
described.

The Algorithm: Detection of Objects
The algorithm is based on global threshold (average) using PSO. To obtain the desired results, the

Page 4

Er. Harish Kundra, Dr. V.K. Panchal, Karandeep Singh, Himashu Kaura & Sagar Arora
International Journal of Engineering (IJE), Volume (4): Issue (1) 89
histograms of several of the scenes in the database containing Objects, were analyzed. From the
analysis, we divided the histogram in five main regions in the basis of Unsupervised learning.
Figure drawn below shows these regions. Region I start from the lowest intensity values to half of
the midway of average intensity. For most of the cases, the intensity values on this region cover
water bodies. Region II covers intensities that go from approximately half of the midway of the
average intensity to midway of average intensity, and typically covers trees. Region III goes from
midway of average intensity, to average intensity and covers bright gray objects such as highways.
Region IV goes approximately from average intensity to midway of highest intensity and generally
covers bright objects like buildings, road dividers, etc. Region V goes approximately from midway
of highest intensity to highest intensity and generally covers barren land(6).


Figure 2: Google’s Earth Image

Figure 3: Histogram of the Image (2)



Morphological Operations
Morphology is a technique of image processing based on shapes. The value of each pixel in the
output image is based on a comparison of the corresponding pixel in the input image with its
neighbours. By choosing the size and shape of the neighbourhood, you can construct a
morphological operation that is sensitive to specific shapes in the input image.

4. SECTION 3: IMPLEMENTATION OF ALGORITHMS TO GET THE DESIRED
RESULT

Computing the Threshold T using PSO
To evaluate the threshold values, we created ten agents which scan their respective regions
assigned to them from the image. Every agent scans the image row wise to find local best values
for each row and eventually evaluates its global best for the assigned region. Among these
regional global bests the agents select the most suitable candidate solution by communicating with
each other. This is known as the Threshold value or the Global best.

Page 5

Er. Harish Kundra, Dr. V.K. Panchal, Karandeep Singh, Himashu Kaura & Sagar Arora
International Journal of Engineering (IJE), Volume (4): Issue (1) 90


Pseudo code for Object Extraction
The process comprises of following steps:
1. Preprocessing: e.g. Atmospheric, correction, noise suppression, band ration,
principal component analysis etc.
2. Selection of the particular feature which best describes the object.
3. Decision: Choice of suitable fitness function for comparing the image patterns
with the target patterns.
4. Assessing the accuracy of the object extraction








Pre-Processing
(unsupervised learning)






































Image Enhancement
Manual interpretation
Decision and Object extract
Extraction of objects
Post-Processing operation
Selection of
threshold
Google Earth’s
image

Feature Extraction
Natural or man-made object extracted