Automatic segmentation of asphalt cracks on highways after large-scale and severe earthquakes with deep learning-based approaches

The detection of land surface cracks caused by earthquakes called seismic fractures is of great importance in terms of finding hidden faults, monitoring their movements and predicting possible earthquakes. On February 6, 2023, there were two devastating earthquakes with the epicenter of Bazaar (Kahramanmaraş-MW=7.7) and Elbistan (Kahramanmaraş-MW=7.6). Earthquakes occurred at 04:17 (01:17 GMT) and 13:24 (10:24 GMT), respectively. These earthquakes occurred on the left-hand lateral-pulsed East Anatolian fault (DAF), one of the two major active fault systems in Turkey. According to the AFAD data, these earthquakes are shallow earthquakes and the first earthquake occurred at a depth of 8.6 km, while the second earthquake occurred at a depth of 7 km. The earthquake was seriously felt in 10 cities, especially the epicenter, and a large amount of building stock was damaged. After the main shocks that occurred, about 09.02.2023 earthquakes were recorded until 16.00h at 1300:00. After these earthquakes, search and rescue teams and relief supplies for 10 provinces need to be delivered to disaster areas as quickly as possible. However, after the Bazaçik and Elbistan earthquakes, the deformations on the roads and airport runways caused a large amount of search and rescue teams and the transportation of relief equipment to the evening. This revealed how important the road quality should be in areas with high earthquake risk. For this reason, deformations of the road network (dense scattered cracks) in earthquake regions should be examined. It is of great importance that the deformation data obtained as a result of these examinations are sorted by expert researchers in the field. The determination of road deformations, especially not of a large degree, will prevent the formation of major deformations due to the effects of the next aftershocks, weather events and traffic that will occur on the roads. For this reason, with the research team to be created, all the deformations that occur on the roads in the disaster area will be examined in detail and the deformation images will be labeled in pixel size. With this dataset, detailed maps of the deformable highways in disaster areas will be made. According to this map, the maintenance-repair plan of deformations that may occur after the earthquake will be able to be prepared. Thus, the reliability of maintenance and repair work will be increased. In addition, the resulting pixel-tagged images will create material for artificial intelligence applications that divide automatic road deformations. With this dataset, the dissertation will provide a deep learning-based approach that automatically divides road deformations.