Extracting the changes of the urban edge area using the classification and segmentation of satellite images with high resolution based on the analysis in geographic information systems.

Master's Thesis in Mapping Engineering

The Trend of Remote Sensing

Abstract

Updating land databases in an urban environment is a difficult and costly task. Satellite remote sensing techniques have been widely used in the extraction and control of land cover change at different scales, which have led to useful results. Doing this becomes easier with the help of automatic change extraction methods. On the other hand, there are two strategies for automatically extracting changes: comparing image with map and comparing image with image. Change extraction methods are mostly based on image to image comparison. In comparing the image with the map, the existing map is used to find the changed areas in the recently acquired image. In order to extract changes, two methods, pixel-based and object-based, can be used. Pixel-based techniques are traditional methods of image analysis, and their working process is to search for differences in various complications. The desired thematic information is extracted from these differences. In many applications, there is a need to extract features that consist of multiple pixels, such as roads, buildings, etc. To extract these complications, group classification of pixels close to each other is needed. Actually, instead of pixel extraction, object extraction is needed here. Objects can have hundreds of features, such as shape, size, spectral response, etc., which can be used for image analysis. Objects have a number of properties that can only be used in object-based methods. In order to extract unknown complications, additional information such as the shape and size of shadows, which can be found in objects, should be used. In this research, a new method based on object-based methods is presented to automatically extract the changes of buildings in urban environments from images with high resolution and using the existing ground database. The desired method consists of several steps. At first, the existing ground database of buildings and other urban objects are modeled. Then the image is segmented. Next, the segmented image is analyzed using the existing database in order to identify the location of the parts where the building has probably changed. Finally, the change extraction rules are tested on the determined parts, and in this way, the parts that present the changed buildings are determined. In the second part of this study, different classification methods have been used instead of image segmentation. On the other hand, in the implementation of the algorithm, features that are suitable for the region and lead to acceptable results should be used. In this study, different geometrical, textural and spectral features were used. The proposed algorithm was tested in three areas. The first area contained 15 objects and not many changes had occurred in this area. The second zone contained 7 objects and was free of change. And finally, in the third area, there were 36 objects, which had more changes than the previous two areas. The result of the proposed method was acceptable in all three regions and most of the changes that occurred in these regions were extracted by this method. The proposed algorithm also has some weaknesses. This method has limitations in extracting the exact distance between buildings. On the other hand, a number of old buildings were identified as new buildings by this method. Another point that should be mentioned is the context of the study area. The data used here are related to the southern edge of Isfahan city. Buildings are visible in the form of completely continuous blocks in this area. On the other hand, the type of satellite image that is used for the operation is also important. The spatial resolution of the image plays an important role in extracting changes. In this research, a Quickbird image obtained in 2008 as well as the city map of 1375/2000 were used.

Keywords: extraction of urban changes, remote sensing, image classification and segmentation, geographic information systems, base law.

Chapter 1

Introduction

1-1- Research problem

Urbanization is an inevitable process that occurs as a result of rapid population growth and economic and social development.The presence of humans in densely populated urban areas, especially in developing countries, causes these areas to be constantly changing. On the other hand, possible natural disasters such as floods and earthquakes also change the face of urban areas [16].

In order to overcome the problems caused by this change and proper urban planning, it is necessary to have up-to-date maps of these areas available. In this way, the main issue in urban planning is the urgent need for updated maps. Due to the rapid and extensive changes in urban areas, it is necessary to pay attention to the change of land covers and uses. Also, according to how the city expands, it is necessary to predict future changes and make basic decisions and urban planning based on these changes[1].

In the past decades, examining the changes in land cover and use has been an important part of global studies. These changes have played an essential role in changing the environment, climate, ecosystem and human activities[3].

These changes are the main factor in changing the world due to ecosystem processes, chemical cycles and most importantly human activities. For this reason, the change of land cover and land use has been discussed as an important project in the programs of international organizations. In the past years, more attention has been paid to the change of urban land cover and use. This is due to the fact that ecosystems in urban areas are more affected by human activities [21] .

The issue of updating the map of urban areas in the whole world is a necessary and obligatory thing that is of great importance for urban planning [30]. The extraction of changes is of great importance in traditional applications such as cartography and new applications such as urban planning and computer graphics. Due to the limitation of natural resources, the processes of updating and producing the land database must be implemented quickly [9]. In order to identify the process of human activities such as industrial development and the conversion of barren lands to arable lands, it is very important to know the changes in land cover and land use during different periods of time. On the other hand, awareness of these changes means that a coded program will be available to achieve a strong economic growth and thus social welfare will increase in the coming years [2]. Recent advances in remote sensing technology and increased access to satellite images with high resolution have created a great potential in determining and displaying a large part of environmental problems in urban areas, and in this way the coded program mentioned above can be implemented more quickly and accurately [6].

The satellite remote sensing method has been widely used in the extraction and control of land cover change at different scales, which has led to the achievement of useful results. This is because remote sensing provides spatial data sets that cover large areas with appropriate spatial details and time interval[8]. Although the extraction of geometric information from images has accelerated in remote sensing in the last decade, the field of photogrammetry has been used to collect information in this field for several decades. In many cases, the goal of photogrammetry has been to update the map from large-scale aerial images [26]. In the last decade, this goal has changed. In this way, photogrammetry is now used to update a GeoSpatial database (spatial or locator) based on scanned aerial images or aerial images obtained directly from digital sensors, and the nature of the collected data is different from old cartographic data [8]. Picture [29]. In the past decades, multi-spectral sensors have been designed for remote sensing that can use the energy reflected from various objects on Earth in the visible and near-infrared wavelengths of the electromagnetic spectrum. Some sensors provide data in the thermal spectrum. While most of today's commercial sensors store data in the visible and near-infrared regions[5].

Updating urban maps is a difficult and costly task. This difficult process can be done by comparing a satellite image of the current situation and the existing map of the region or comparing two images at two different times in a simpler way [28].