Obje tabanlı sınıflandırma yaklaşımı ile 3. seviye ulusal arazi örtüsü/kullanımının belirlenmesi

Fen Bilimleri Enstitüsü, Geomatik Mühendisliği Programı, İTÜ

Obje tabanlı sınıflandırma yaklaşımı ile 3. seviye ulusal arazi örtüsü/kullanımının belirlenmesi

Ş. Özge Dönmez; Elif Sertel, 2015

Abstract: Natural resources on the Earth are limited and resources is rapidly declining with each passing day. Therefore, both the academic studies and the societal level land management is becoming more important every day in terms of follow-up issues of natural resource use. Natural or caused, such great losses occurring from human-induced fires in forest areas, wetlands, and changes in agricultural land is only one reason to turn into artificial field of natural areas. Similarly, all these changes, production plan, when considered together with the right environment and the negative effects of the increasing population and urbanization is increasing the importance of management in a planned manner. Thus, many technical and scientific work is used in an expeditious manner for further analysis in this subject. Land cover consist of built-up areas, agricultural areas, forests and semi-natural areas, wetlands and water bodies formed by the physical and biological cover of the surface. In Turkey, there is no official standard method of management for the land cover data. On the other hand, land cover data used by many ministries and different department of the general directorates in our country. Land use and land cover maps can be useful in the monitoring of urban development,management of natural resources, protection of wildlife, as a base for geographical information systems, damage assessment (hurricanes, floods, fires, volcanic eruptions and earthquakes), determining of legal limits for property assessment and tax collection, etc. Thus, forming an up-to-date land cover map is a crucial matter. It is easily possible to form up-to-date land cover maps with speed and accuracy using remote sensing technology. Via land cover change analysis applications that will be built with the help of remote sensing, it is possible to accurately predict forest vegetation changes, losses and capacity of regeneration; forest fires, transformation of wetlands, topographic and urban transformations and changes in agricultural patterns. Applications that utilize remote sensing technology are preferred in this project as they are the leaders of the new age. Not only that, but remote sensing data is also easily updateable, has high spatial resolution to study on, is much more budget and time-friendly, and provides opportunities for areas that are not immediately accessible for surface work. Hence, the utilization of remote sensing data and technologies are planned methods to create the land cover maps in this Thesis. Around the globe, studies on this issue have started developing years ago and are steadily increasing in importance. In this context, the Central European Committee (CEC) launched the Project ‘CORINE-Coordination of Information on the Environment’ many years ago. Project CORINE was lead by CEC from 1985 to 1990, its terminology and methodology developed. In the past, land consolidation projects only covered rural and urban development and areas affected by changes in basic infrastructure. However, with the CORINE land cover program, the integrity of the European Union states’ environmental data was collected and policies regarding the acquired information were created. Similarly, the United States has land cover systems formed upon the individual standards of its states and cities. Immediate examples would be various systems developed by environmental corporations: USGS (United States Geological Survey), CORINE, FAO (Food and Agriculture Organization) etc. Last of all, several studies about these are grown and developed rapidly and remarkably. In that case, whole European region’s environmental inventory has been reached with CORINE Program, and developed and assigned valid strategies about now and future. CORINE Program is the biggest land monitoring project in the Europe. In 38 countries, a total of 5.8 million km2 of area are monitored , and for three different times has completed (1990, 2000, 2006) and 2012 program is still ongoing for some countries includes Turkey. In this project, 1 : 100000 map scale is used, additionally minimum mapping unit standard is 25 ha, but for change presentation this standard is defined as 5 ha. CORINE Nomenclature includes 3 hierarchical levels within 44 land cover and land use classes. Level 1 classes are artificial surfaces, agricultural areas, forest and semi-natural areas, wetlands and water bodies. Level 2 has 15 classes. Four of it are related to artificial surfaces, four of it are related to agricultural areas, three of it are under forest and semi natural areas, two of it are under wetlands and the last two classes in level 2 are related to water bodies level 1 class. Istanbul, one of the most important cities since Turkey has been elected as a study area. Within the borders of Istanbul in a predetermined area, after the image processing phase from the SPOT 6 Satellite Image 1.5 m resolution (2013) and auxiliary data, object-based classification is conducted; and finally level 3 land cover/land use map is created. Object based classification is required to work on objects, not pixels. After the image processing phase, the segmentation phase came and choosing the proper parameters, two different segmentation phases were applied. By this way, the smallest unit of the image, turned to a segment from a pixel by segmentation process. In application, spectral and spatial functions are used respectively, and a decision tree is created for object based classification. Also, screen aided visual interpretation keys are used simultaneously besides automatic classification steps. In this manner, fuzzy classification algorithms are produced by aid of decision tree, various functions, spectral indices, band ratios, band combinations, determining optimal value intervals etc. in order to produce land cover and land use map of İstanbul. As the study area, İstanbul – one of the most important cities of the country – was chosen. In the Period of accession process to the European Union, the pilot study was vital for improving land cover and land use classification system that was compatible with CORINE project with using high resolution remotely sensed data. In application, image processing steps are performed that initially. To fulfill the aim, object-based classification was conducted, and finally the land cover/land use map was created in 1 : 25000 scale and with 1.56 minimum mapping unit standard. As a result of classification, accuracy assessment has been performed. As a result of the classification; total areas are calculated separately for each land use and land cover classes. These kind of studies – land cover and land use producing and updating – are important and pioneer for future planning in national and global scales in order to protect the limited natural resources of Earth and the optimal management of habitat.

Tez No: 397932

İlginizi Çekebilir

Vessel Detection From Very High-Resolution Satellite Images With Deep Learning Methods

İletişim Sistemleri Anabilim Dalı, Uydu Haberleşmesi ve Uzaktan Algılama Programı, İTÜ Vessel Detection From Very …

CSCRS