Page 700 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

Object Oriented Post-Classification Land Use/ Cover Change Detection on the Mehr
Pardis Housing Construction Area

1
Somayeh Bayat , Fatemeh Tabib Mahmoudi 2*

1 First Department, First University, Address, City, Country Name
2
Second Company Department, Company Address, City, Country Name
* fmahmoudi@sru.ac.ir

Abstract: The increasing population of the large cities has led to developing new constructions in areas around cities to
create settlements for the overflow of the population. In this paper, the impact of Mehr Pardis housing construction is
investigated on the land use/cover changes. For this reason, Landsat satellite images have been used in 17 years’ time
interval, between 2002 and 2019. After performing initial image processing and image segmentation, the three object
classes of residential buildings, vegetation, and soil were identified by the object based image analysis procedure. Then,
post-classification change detection performed on the generated object based classification maps of both 2002 and 2019
epochs. The obtained results of this study represent a 184% increase in the number of buildings and structures
constructed in this urban area.

1. Introduction
Changing the land covers to facilitate the human life is among the most important changes in land surface that have
significant effects on the natural resources and environmental processes. With the rapid urbanization, population growth in
cities and consequently urban expansion, environmental and natural coverage of areas around metropolises such as Tehran are
changed to prepare for the overflow of urban population. Such changes in natural land cover not only disrupt the thermal
balance, but also have negative effects on the landscape, energy efficiency, health and quality of human life. Therefore, it is
important for planners and city managers to be aware of the land use/ cover changes, especially in metropolitan areas during
long-term periods, in order to assess and anticipate the resulting problems.
For spatial information acquisition based on monitoring the land use/ cover changes in urban areas, various sources of
remote sensing data have been widely used [1-3]. Multi-temporal remote sensing data are one the powerful tools for detecting
land use/cover changes due to the increasing urban growth and then, for updating the three dimensional city models [2]. Land
use/cover changes detection methods using various types of remotely sensed data have been proposed by many researchers to
meet a wide range of applications [3]. Considering the procedure and the utilized multi-temporal remote sensing data, change
detection algorithms can be divided into two dimensional and three dimensional categories [4]. Some of the proposed land
use/cover change detection methodologies have utilized only the multi-spectral remote sensing data without considering digital
elevation models, which is applied in the absence of elevation data [5-8].
In some researches, two-dimensional change detection methods are classified into pixel-based and object-based
algorithms considering the computational basis. Pixel-based methods are also divided into the direct comparison,
transformations, post-classification methods, machine learning and advanced methods. Object oriented methods are also
categorized into direct object comparison and object-based post- classification [9].
Despite the relative weakness of pixel-based methods compared with object-based ones due to only utilization of
spectral features in change analysis, simplicity, high diversity of algorithms and their high levels of automation have made
these techniques widely developing for change detection based on airborne and space borne remote sensing data [10].
In object-based methods usually a segmentation process is performed for separating the homogenous image objects.
Image segments contain more than one pixel, so object-based image analysis methods are utilized geometric information
together with spectral features for classification of the image segments. In object-based change detection methods, a direct
comparison of the multi-temporal classified objects is performed. In this method, the comparison is based on the geometric
characteristics and spectral features of the image segments [10, 11].

2. Study Area and Data Sets
The study area in this research is the new city of Pardis located in the center of Pardis province, 17 km northeast of
Tehran. According to the 2016 census of Iran, the population of this city was 73,363 people. Meanwhile, the population of
Pardis city in 2006 was equal to 25,360 people and in 2011 was equal to 37,257 people. The reason for this increase in

1