Forest stand delineation using Ikonos image and object based image analysis

Forest stand delineation using Ikonos image and object based image analysis Abstract: Together with the developments in satellite technology, it is considered that high resolution satellite data may be used as an alternative source of information to aerial photos in delineation of stand types. The study aims to reveal how detailed one could work to generate the map of stand types which form the basis of forest management plans using IKONOS satellite data. For this purpose, object based classification was applied to satellite image. Firstly, image segments which represent target objects were generated applying image segmentation algorithm to the satellite image. The image segments generated at three different levels according to different scale parameters and homogeneity criteria were classified according to standard nearest-neighbor approach. Classification accuracy was determined using both the stand maps of study area and ground control points. Overall accuracy was calculated as 58% (Kappa=0.54). Accordingly, it was understood that it was not possible to generate a stand map with sufficient accuracy from the IKONOS satellite image using automatic classification. Keywords: Ikonos, forest inventory, image segmentation, object based classification, stand map Ikonos goruntusu ve obje bazli goruntu analizi kullanilarak mescere tiplerinin ayrilmasi Ozet: Uydu teknolojisindeki gelismelerle birlikte yuksek cozunurluklu uydu verilerinin, mescere tipleri ayriminda hava fotograflarinin yerine alternatif bir bilgi kaynagi olarak kullanilabilecegi dusunulmektedir. Calismada, IKONOS uydu verisinden amenajman planlarinin temelini olusturan mescere tipleri haritasini duzenlemek icin ne kadar ayrintiya gidilebileceginin ortaya konulmasi amaclanmistir. Bunun icin uydu goruntusune obje bazli siniflandirma islemi uygulanmistir. Uydu goruntusune oncelikle goruntu dilimleme islmei uygulanarak, hedef objeleri temsil edecek goruntu dilimleri olusturulmustur. Farkli olcek parametreleri ve homojenlik kriterlerine gore uc farkli seviyede olusturulan goruntu dilimleri, standart en yakin komsu yaklasimina gore siniflandirilmistir. Siniflandirma sonuclarinin dogruluk degerlendirmesi calisma alanina ait mescere tipleri haritasindan ve arazi calismalari sirasinda alinan denetim noktalarindan faydalanilarak yapilmistir. Mescere tipleri duzeyinde yapilan siniflandirma sonuclarinin toplam dogruluk degeri %55 (Kappa=0.52) olarak hesaplanmistir. Buna gore, IKONOS uydu goruntusunden otomatik siniflandirma ile yeterli dogrulukta mescere tipleri haritasinin uretilmesinin mumkun olmadigi anlasilmistir. Anahtar Kelimeler: Ikonos, orman envanteri, goruntu dilimleme, obje bazli siniflandirma, mescere haritasi Received (Gelis):  11.01.2016 - Revised (Duzeltme):  18.01.2016 - Accepted (Kabul):  22.01.2016 Cite (Atif):  Ozkan, U.Y., Yesil, A., 2016. Forest stand delineation using Ikonos image and object based image analysis. Journal of the Faculty of Forestry Istanbul University 66(2): xxx-xxx. DOI: 10.17099/jffiu.xxxxx