Özet:
This thesis describes a maritime geometric collision probability model which can be used in congested narrow waterways. Geometric collision probability model is extended and applied to a spatially distributed maritime tra c, rather than the route based models. It nds out the hotspots for the collision places on a spatial representation. Model is applied to the Strait of Istanbul with automatic vessel tracking data. Results of the model are in line with the past events and theoretical studies. The aim of the study is to quantify the geometric collision probability and represent the results on probability map. A long-term maritime tra c data is collected from February 2014 to August 2016. The data is parsed and stored in a SQL database system. Size of the database is 94 gigabyte. Grid based analysis method is used for the big data management. Geometric collision method is applied with the output of this database. There are two main part of this study. One of them is the detailed analysis and modeling of maritime tra c. The other one is the geometric collision model application. Maritime tra c analysis is based on automatic vessel tracking data. Ship characteristics, navigation patterns and tra c densities along the Strait are given in the analysis. Analysis shows the captains route preferences along the Strait and speed according to their travel directions. North entry ships have a tendency to travel faster than the other directions. Also one more important nding is the net transfer of goods direction in the Strait is from north to south. Geometric collision model gives the probability map of the Strait in spatially distributed way. According to the model the highest collision probable sector is between Sariyer and Umurbey (Sector 10) and the highest probability is found between Kandilli and Rumelihisari.
