ÇOK SPEKTRUMLU VERİLERDEN BİLGİ ÇIKARIMINDA MEKANSAL FİLTRELEME ETKİSİNİN İNCELENMESİ (INVESTIGATING THE EFFECTS OF SPATIAL FILTERING OF INFORMATION EXTRACTION ON MULTISPECTRAL IMAGERY)

INVESTIGATING THE EFFECTS OF SPATIAL FILTERING OF INFORMATION EXTRACTION ON MULTISPECTRAL IMAGERY Determining the land cover and land use changes are important for monitoring and managing the changes and their causes in both natural environment and man-made areas. Quantitative analysis based on numerical computations and visual interpretation based on human eye are the techniques used in information extraction -related to land use and land coverfrom digital imagery. The quality of visual interpretation depends on enhancing the difference between land characteristics on the imagery as much as the performance of the analyst. Spatial filtering is one of the spatial characteristics manipulation techniques for information extraction and it surpass or emphasize some image characteristics. This study was conducted in Tekirdağ, Lake Bıyıkali and its neighborhood (area of 2500 hectares) where is rich in landcover and land use. The study aims to extract information from imagery by filtering techniques and therefore it was carried out as two stages. Firstly, different spatial filters were applied to Landsat5 TM imagery -which has medium spatial resolutionin order to determine the filter providing the best distinguishability, compare it with other filters and define the land use and land cover areas. The spatial filters used in the study are; 3x3 sized high pass, edge enhanced , Laplacian edge enhance and high frequency components enhanced filters. Secondly, some land use/land cover objects on enhanced Landsat5 images were digitized and areal comparison was conducted by using Worldview-2 image as a reference. Finally, the relative root mean square errors (RRMSEs) of the areal differences acquired from both reference and enhanced images were calculated in order to assess the accuracy and the best filtering method/methods were determined.