Image Analysis of Measuring Building Configuration for Seismic Damage Estimation

Abstract

A novel methodology of image analysis was designed to assess more quantitative seismic damage estimation at the building level for a city area in which 10–30% of the entire building stock was damaged. The methodology is capable of accurate configuration determination, which forms the kernel part toward seismic damage estimation based on displacement detection. This provides the following advantages. (1) A narrow stripe is extracted around an approximate building configuration in a virtual image generated based on high-resolution geographical information system (GIS) data. Analysis of building edges incorporated by the narrow stripe reduces computation cost and helps better distinguish building edges from background edges, such as trees, shadows, and windows. (2) An edge detection technique combined with a cumulative Gaussian distribution method and a gradient descent searching algorithm is validated. It offers subpixel location accuracy of a configured structure in a two-dimensional (2D) digital image. (3) Given the approximate position, orientation, resolution, view angle, screen distance, and upward pointing vector in the camera coordinate system, a three-dimensional (3D) wireframe city can be presented pre- and postearthquake.