Detecting Urban Vegetation from Different Images Using an Object-Based Approach in Bartin, Turkey

Urban vegetation plays an important role for sustainable development policies, environmental conservation and urban planning process of a city. It is necessary to detect the amount of green areas and their distribution to form the ecosystem model of the urban environment. It is quite important to use satellite imagery having different ground sampling distance (GSD) in the economic and accurate detection of urban green areas. Especially object-based image analysis has been frequently used today for object extraction processes. In object-oriented image analysis, not only pixel gray values but also spectral and contextual data that help to distinguish the segments consisting of interrelated pixels on the image are used. For this reason, more positive results are obtained in comparison with pixel-based approaches. In this study, city center of Bartin, in which there is a rich amount of green areas and its vicinity was chosen as the test area. As the satellite image data, LANDSAT 7 ETM+ (28.5 m GSD), SPOT 4 Level 2 A (20 m GSD) and IKONOS (1 m GSD) were used. Test area was divided into segments involving lots of different classes on each image. Urban vegetation class was formed by determining the suitable functions for the objects which will be involved in the urban vegetation class. eCognition v4.06 software was used for object-based classification analysis. Classification results were transformed into vector data and visual and digital analyses were made using GIS.