"Earthquake damage assessment using objective image segmentation: A cas" by Thomas Oommen, Umaa Rebbapragada et al.
 

Earthquake damage assessment using objective image segmentation: A case study of 2010 Haiti earthquake

Document Type

Conference Proceeding

Publication Date

12-1-2012

Abstract

The free and open availability of high-resolution satellite and airborne imagery after the 2010 Haiti earthquake provided a new opportunity to analyze earthquake-induced damage. Previous work has analyzed pre- and post-event remote sensing images using pixel-based change detection. The drawback of these methods is that they have difficulty differentiating earthquake-induced damage from other natural and manmade changes. An alternate method is to analyze damage in terms of objects rather than image pixels. The significant advantage of object-based analysis is its ability to effectively use the characteristics of the change such as shape, length/width ratio, asymmetry etc. to differentiate between damage types. However, to make effective use of object-based analysis for earthquake-induced damage evaluation, we need a comprehensive understanding of all potentially useful earthquake-induced damage characteristics, specifically from a segmentation-based perspective. In this study, we perform a case study on imagery from the Haiti earthquake that evaluates a novel object-based approach for characterizing earthquake induced surface effects of liquefaction against a traditional pixel based change technique. Our technique, which combines object-oriented change detection with discriminant/categorical functions, shows the power of distinguishing earthquakeinduced surface effects from changes in buildings using the object properties concavity, convexity, orthogonality and rectangularity. Our results suggest that object-based analysis holds promise in automatically extracting earthquake-induced damages from high-resolution aerial/satellite imagery. © 2012 American Society of Civil Engineers.

Publication Title

Geotechnical Special Publication

Plum Print visual indicator of research metrics
PlumX Metrics
  • Citations
    • Citation Indexes: 5
  • Usage
    • Abstract Views: 4
  • Captures
    • Readers: 14
see details

Share

COinS