Comparison of Pixel Based And Object Based Classification Methods on Wetland Areas: Example of Aslantaş Dam Lake
Mustafa Hayri Kesikoğlu Sevim YASEMİN ÇİÇEKLİ Tolga Kaynak
AbstractBy the development of technology, image classification algorithms frequently use to identify land use and land cover of any area in remote sensing studies. Due to the diversity and complexity of land cover on the wetland areas, it is quite difficult to obtain accurate results related to the earth's surface. The main purpose of this research is to compare the overall accuracies of object based and pixel based image classification methods. Arslantaş Dam Lake is structured on Ceyhan River for irrigation, flood control and electricity generation in Osmaniye province. In this study, Landsat-8 LDCM satellite image of Aslantaş Dam Lake with spatial resolution of 30m, acquired on December 29, 2017 was used. Firstly, image was classified by pixel based classification with support vector machines (SVM) method. After that, image was reclassified by object based classification with K-nearest neighbour (KNN) method. Five classes namely lake, agricultural area, soil, vegetation and building area were determined by using these algorithms. Ground truth data were gathered from aerial photographs, available maps and personal informations. Finally, overall accuracies of these methods were compared. It is observed from the classification results that object based KNN method provide higher accuracy than the other classification method.