Dedection of shallow landslides with object based classification approach from unmanned aerial vehicle data

The Black Sea Region is one of the most landslide prone area due to the high slope topography, heavy rainfall and highly weathered hillslope material conditions in Turkey. Generating of landslide inventory maps are first step producing landslide susceptibility maps of the region. Terrestrial methods for mapping the landslides occurrence to the region are time consuming and costly due to the conditions of the region. Landslide mapping based on satellite images and aerial photographs taken from airplanes has some limitation factors such as climatic conditions, cost and limited repetitive measurement capacities. In addition, visual interpretation-based landslide mapping, which is based on satellite images and aerial photographs, is time consuming. Therefore, the data collection based on Unmanned Aerial Vehicle (UAV) and generation of landslide event inventory maps by object-based classification approach through this data can be superior in terms of speed and cost compared to other methods. In this thesis study, a semi-automatic model was developed by object-based classification approach for rapid mapping of landslides from data obtained from unmanned aerial vehicles after major landslide events in the Black Sea Region. Within this scope, two research areas were selected, namely Bartın - Kurucaşile and Rize – Çayeli. Landslide mapping Models have been developed in the selected areas and the successes of the models have been investigated in the test areas. The landslides obtained with the developed models have been compared with the landslides produced by the experts in numerical and area based. As a result of the comparison process, landslides can be mapped with an accuracy rate of 86.27% according to the number of landslides and 83.01% according to the landslides area.