AUTOMATIC IDENTIFICATION OF BUILDING FEATURES FOR SEISMIC DAMAGE ASSESSMENT ON A LARGE SCALE

Seismic risk assessment represents a big challenge in countries with a high seismic hazard and a significantly vulnerable built heritage, such as Italy. When carrying out seismic risk evaluations at large scales, however, the identification of buildings and their features can be very costly and time consuming. In this work, artificial intelligence techniques are used to automatically and remotely retrieve exposure information. First, building data are collected using satellite imagery, then street-view pictures are extracted for each building and Convolutional Neural Networks are trained to recognize specific features of interest, especially those that affect seismic vulnerability. Furthermore, a seismic damage calculation platform is developed. The results provided by this algorithm can be useful for managing emergencies and establishing priority criteria for seismic mitigation strategies.