describe the different types of settlements is insufficient due to 
variation in the structure, material, shape and so on. Hence 
more refined methods such as object oriented approaches are 
necessary to detect the informal settlements from very high 
spatial resolution data. 
With the wide availability of VHR images, automatic object 
delineation techniques are being extensively researched and 
have proven to be accurate in urban applications (Ebert, Kerle, 
& Stein, 2009; Sliuzas, Kerle, & Kuffer, 2008). Recent 
developments in *object-oriented" image classification (based 
on image segmentation) have taken advantage of the detailed 
spatial characteristics of high-resolution datasets. The research 
in this area has emphasized the reduction of spectral variability 
within the objects and the incorporation of additional 
information from spatial and contextual image/ object 
characteristics (Johnsson, 1994; Blaschke and Strobl, 2001). 
OOA is capable of using multiple data types during analysis to 
create meaningful segments. Segmentation 1s an important step 
preceding the classification of image objects. The classification 
process can include a variety of information, ranging from 
spectral mean values for each object, to measures of texture, 
context and shape. OOA offers great potential because of its 
ability to include spatial, spectral and contextual characteristics 
similar to human cognitive image interpretation (Hofmann, 
2001; Herold et al, 2002; Van Der Sande et al., 2003; Benz et 
al., 2004). 
Even though several studies showed the feasibility of detecting 
slums by using OOA, their relatively high inner-structural 
heterogeneity and their varying pattern impede the generation 
of an automated detection process. In the present paper, the 
visual interpretation indicators used by experts for slum 
identification and ground knowledge of slums in the city of 
Pune, India were used to semi-automate the classification in an 
OOA environment 
2. MATERIALS AND METHODS 
2.1 Study area 
Pune the city (Figurel), selected for present study is one of the 
fast developing urban agglomerations in Asia. 
india Rl Maharashtra ^ PuneDiatrier 
  
  
  
  
  
  
  
Figure 1.Location of Pune city and Quick Bird scene 
  
  
  
It ranks eighth at national level and second at state of 
Maharashtra (Census of India, 2011). It has grown manifolds 
over the past two decades in terms of population and area. 
Pune city lies between latitudes 18?25"N and 18?37'N and 
longitudes between 73°44°E and 73° 57°E and cover an area of 
243.96 sq.km. Between 1976 and 1981 the population of Pune 
city (Table 1) grew by 16.7%, from 1981 to 1991 it grew by 
30.2%, and between 1991 and 2001 growth increased to 
62.17% (Shekhar., 2004;Shekhar.,2006).The recent census 
2011 also showed decade growth rate of 22.6% (Census of 
India, 2011). India’s Town and Country Planning Organization 
(TCPO), the technical arm of the ministry of urban 
development, ranks Pune as a city with third largest number of 
slums in India. 
  
  
  
  
  
  
  
  
  
Census year Population 
1951 400902 
1961 794052 
1971 1029466 
1981 1202848 
1991 1566651 
2001 2540069 
2011 3115431 
  
  
  
Source: Census of India 
Table 1. Population of Pune city 
Environment Status Report of Pune city for 2009-10 had stated 
that at the rate at which the slums in the city were growing, at 
least 50 per cent of the citys population would be living in 
slums. Pune slums are congested, have structures made of 
materials which are considered garbage, such as wood used for 
packing, plastic sheets, opened out metal tins, galvanised iron 
sheets, bamboo sheets, etc. and often lack the most basic of 
facilities for all its inhabitants. Pune's slum population is 
scattered across the whole city. 
2.2 Data base 
As data source Quick bird scene (60 cm spatial resolution) was 
used showing parts of Pune city (Figure 1).The image shows 
the central part of Pune city comprising typical urban features 
including slums and other formal areas. Cloud free Pan 
sharpened data with three bands (RGB) of the year 2006 was 
used for detecting slums from non-slum areas. 
GIS layers of road and water bodies were also used as thematic 
layers in the object oriented analysis. Census data and 
Environmental Status reports of Pune city Municipal 
Corporation were used as secondary data for this study. 
2.3 Methodology 
The first step towards identifying the slums in OOA is to 
generate segments, i.e. an automatic division of an image into 
coherent groups of pixels (segments, objects) and the criteria 
used to segment an image is a degree of homogeneity within 
each particular object and heterogeneity among neighbouring 
objects (Baatz. M., A.Schape, 2000). It was done by using 
Multi resolution segmentation with the objective choice by 
using Estimating Scale Parameter (ESP) tool (Drágut et al., 
2010). These outputs are called *object primitives', which lead 
to meaningful ‘objects of interest’ by further refinement 
(e Cognition, 2010). The segmentation used all image layers as 
well a 
meanir 
level, 
and en: 
in an 
enhanc 
As th 
classifi 
unders 
identifi 
Develo 
process 
extract 
classifi 
class 
descrip 
conditi 
assigne 
propert 
Shado 
Figure 
3.1 Se 
In orde 
units, t 
comple 
means: 
the mo 
criteria 
split in 
down): 
basic « 
method 
pixels 
contras 
comple 
clusters 
heterog 
approac