Application of Object Based Image Classification Using Orthophotos: A Case Study of the Van-Erci？ Earthquake

Earthquakes are the most destructive natural disasters, causing massive loss of life, infrastructure damages and serious financial losses. Earthquake-induced building damage detection is a very important step after earthquakes since such buildings pose a risk for cities. The recent advances in digital photogrammetry/remote sensing technology and satellite sensor/camera systems have inspired to emerge new and different research areas and applications. With the development of digital camera technology since 1950's, it has become possible to produce orthophotos with high spatial resolution and positional accuracy. These orthophotos are used successfully in the extraction of the details and the thematic classification. Depending on the high spatial image resolutions, the use of object-based approach has gained importance and has become widespread rather than the traditional pixel-based approach in the thematic classification. The town of Erci？ in Van province, situated in the eastern part of Turkey, was selected for the study. In this region, an earthquake with moment magnitude (Mw) of 7.2 was occurred at 10:41 UTC (13:41 local time) on October 23th, 2011. In this study, the orthophotos produced from the aerial photographs taken one day after the earthquake by the General Command of Mapping for the purpose of damage detection. In the post-earthquake ortophotos, 2 different types of sites were selected as the study areas. As an initial step, segmentation process was carried out by selecting the most suitable parameters for the classification of both homogenous and heterogeneous areas. In the segmentation phase, the multi-resolution algorithm was utilized. After doing different trials, the most suitable shape factor and compactness values were considered. In the next step, two different classification approaches, namely “supervised” and “unsupervised”, used within the object based image classification were applied and their classification accuracies were compared.