Babak Mansouri - International Institute of Earthquake Engineering and Seismology (IIEES), Tehran
Shakiba Mousavi - International Institute of Earthquake Engineering and Seismology (IIEES), Tehran
Kambod Amini Hosseini - International Institute of Earthquake Engineering and Seismology (IIEES), Tehran

A strong earthquake with Mw ۷.۳ occurred on Nov. ۱۲, ۲۰۱۷, at the Iran-Iraq border where the city of Sarpol-e Zahab (۲۴ km from the epicenter) and many other towns and villages were affected severely. Rapid damage mapping essentially helps to understand the location, the extent and the severity of high hit areas, and it is regarded as an important source of information in assisting proper crisis management. Rapid damage mapping can be completed according to three general methods namely; manual, semi-automated, and automated. This research explores a proposed semi-automated method that once calibrated and operational it can be utilized as an automated process. After preprocessing the satellite data, individual buildings or building parcels are identified using either some building extraction tools or with the use of some ancillary data sets. The proposed damage detection algorithm is based on deriving a set of textural indices associated to individual building or property footprints. These parameters have been input into an Artificial Neural Network (ANN) for damage classification. The trained ANN created urban damage maps. For detecting significant observable physical changes/damages to the buildings, two schemes were developed: ۱) by comparing the post-event with the pre-event VHR satellite images, and ۲) using a post-event image only. In scheme-۱, before and after images were acquired from different satellites (TripleSat_۲ with ۸۹ cm and SV۱ with ۵۰ cm spatial resolution) as input and the Overall Accuracy (OA) of the proposed damage classification was reported as ۷۲%. The damage classification accuracy in scheme-۲ produced an OA of ۷۵%.

Object-oriented image processing, damage detection, Textural Analysis, Artificial neural network