Instant optimization of the shortest routes based on automated detection of the damaged roads after an earthquake with artificial intelligence

After earthquakes, which is one of the natural disasters, identifying damaged roads and finding suitable undamaged roads are of vital importance in carrying out rescue operations, evacuating people from the affected area as soon as possible or dispatching humanitarian aid materials to the region. Road damage assessments made with traditional human observation and satellite images used today are time consuming or costly and do not contain any information about traffic passage in order to reach disaster areas by only examining the damages on the roads. In this study; a fast and efficient hybrid transportation system has been created by integrating the convolutional neural network and the classical shortest path algorithm to automatically determine the road damages that occur after earthquakes and to determine another route close to the disaster areas. The CNN algorithm is trained on a dataset consisting of a total of 552 road images, both damaged and undamaged. After the test performed with the VGG-16 architecture, it managed to achieve 97% accuracy. With the predictions made by the CNN algorithm as 'open road' and 'closed road', the damaged roads over the classical shortest path algorithm were removed with the penalty term used and the route with the shortest distance was determined among all available routes. Successful results were obtained by creating a fast and efficient hybrid transport system with the integration of these two algorithms.