Urban roads damage detection using an ANFIS and SVM classification from post-earthquake QuickBird image (Case Study: BAM, Iran)

Finding damaged roads and obstacles after a devastating earthquake to determine optimal paths and conducting rescuers is essential. In this research, a new semi-automatic ANFIS approach is proposed to detect damaged roads in urban areas by use of pre-event vector map and post-event QuickBird image. In this study, damage is defined as debris caused by collapsed buildings adjacent to the roads. Vector map and GIS data is used to detect road regions and some features are generated on original bands. A support vector machine (SVM) classification is applied to label damaged pixels. Finally, after dividing roads to the equal partitions, an appropriate Adaptive Neuro-Fuzzy Inference System is proposed for detecting damaged partitions. Results show that ANFIS has achieved overall accuracy of 92% in detecting obstacle partitions. This indicates the efficiency and accuracy of the proposed method for road damage and obstacle detection.

کلیدواژه ها:

road damage and obstacle detection ، Adaptive Neuro-Fuzzy Inference System (ANFIS) ، QuickBird images ، Support Vector Machine (SVM)