← Back to Sessions

A Feature-Based Liquefaction Repository From Aerial and Satellite Imagery for Computer Vision Analysis

Session: Application of Remote Sensing and Space-Based Earth Observations Data in Earthquake Research II

Type: Oral

Date: 4/22/2021

Presentation Time: 05:15 PM Pacific

Description: 

Mapping of seismic-induced soil liquefaction is a common practice after earthquakes to help local authorities devise better disaster preparedness plans after a seismic event and to provide data for the research community. Many of those reports provide limited and inconsistent geographic information. Some of the reconnaissance reports provide coordinate-based region locators, and others provide more qualitative, general locations, while some more recent efforts provide digital polygons. The inconsistencies in how the liquefaction field investigations are communicated have led to the need for more uniform datasets.

In addition, with the increase in availability and resolution of both aerial and satellite imagery, there is the potential for automatic detection and mapping of liquefaction after earthquakes using image processing and artificial intelligence algorithms. Therefore, the aim of this research is to build a uniform liquefaction image library or repository from aerial and satellite imagery which will provide digital polygons for individual liquefaction features – sand boils, lateral spreading, and settlement. The liquefaction image library will include labels providing such information as infrastructure impacted and confidence level in the feature type and boundary. To ensure higher accuracy of liquefaction detection, image analysis methods require that the liquefaction features be outlined with polygons and that those polygons be limited to only liquefaction features. We present our preliminary liquefaction image library and demonstrate its use in automated models of liquefaction detection from post-event imagery.

Presenting Author: Christina Sanon

Student Presenter: Yes

Authors