Page 701 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

population can be attributed to the beginning of the construction of Mehr housing in this area, because in recent years, this area
has been the target of construction of towers and apartments in Mehr housing. In this study, the effects of Mehr Pardis housing
constructions have been investigated on the rate of land use/cover changes in this region over a long period of time and based
on using the remote sensing data and image processing techniques. The reason why this area was selected as a case study is
that in the reports and news presented in recent years, the rate of construction and land use/cover changes in this area has been
very high.
In this study, two Landsat satellite images taken from the Pardis area over a period of 17 years have been used. The first
image was taken in 2002 from Landsat-7 ETM+ sensor and the second image was taken from Landsat-8 OLI sensor in 2019.

3. Proposed Change Detection Strategy
The proposed land use/ cover change detection strategy in this paper is a post-classification method based on
performing object based image analysis procedure. First of all, the pre-processing operation is applied on the multi-temporal
images from the study area to prepare them as input to the change detection algorithm. Then, by object based classification of
the multi-temporal images, classification maps are generated. Finally, with comparing and differentiating the multi-temporal
classification maps, the results of detecting environmental changes such as the rate of increase in construction, changes in soil
area and vegetation are obtained. As illustrated in Figure 1, the proposed land use/cover change detection algorithm in this
study consists of three main stages of data pre-processing, multi-temporal object based classification and change detection.

Fig. 1. Proposed land use/ cover change detection strategy

3.1. Data Pre-processing
In the pre-processing stage, radiometric, atmospheric and geometric errors are corrected as much as possible with a
series of operations on remote sensing data. In this study, FLAASH atmospheric correction filter was used to perform
atmospheric corrections [12] (Pietro-Amparan, 2018). There is no need for geometric corrections as the images taken at both
times are geocoded and have UTM global coordinates.
The spectral bands of Landsat-7 and Landsat-8 images have a spatial resolution of 30 meters, and both satellites have a
panchromatic band with a spatial resolution of 15 meters. In this study, to increase the sharpness of the images, pan-sharpened
products were made using the panchromatic band, as a pre-processing step to obtain the better results from object based
classification process.

3.2. Object Based Classification
Object based image analysis procedure is used for image classification in this paper. Two main steps are considered in
performing such analyzes: 1) application of segmentation algorithm on multi-temporal images, 2) knowledge-based
classification of image segments based on the extracted features. The image objects obtained from segmentation algorithm are
used as basic units in knowledge-based classification. Therefore, by using precise methods in segmentation, confidence in the
knowledge-based classification results of the image segments increases [13] (Tabib Mahmoudi, 2014).
In performing object based analysis in this research, the multi-resolution segmentation algorithm has been used to
produce the basic units for knowledge-based classification. The multi-resolution segmentation technique begins with the

2