Automation of Post-Earthquake Civil Infrastructure Reconnaissance

Traditionally post-earthquake structural engineering reconnaissance consists of a team of experts who are deployed to the field to record and capture earthquake damage data, which is later uploaded into online repositories. Despite many advances to these data archives in recent years, the entries in online repositories often have limited metadata which make it difficult and time consuming to extract specific damage evidence that can be used for meaningful analysis. This report outlines the author’s contributions to overcoming these challenges via the development of a neural network that automatically filters and classifies post-earthquake civil infrastructure damage data after a seismic event. In addition, this document summarizes the author’s progress on utilizing neural network output to rapidly generate geospatial maps from reconnaissance data