سال انتشار: ۱۳۸۶

محل انتشار: پنجمین کنفرانس بین المللی زلزله شناسی و مهندسی زلزله

تعداد صفحات: ۷

نویسنده(ها):

Shirzaei – National Center of Earthquake Prediction, IIEES
Mansouri – Center of Excellence in Risk Management, IIEES

چکیده:

Remote sensing technology has shown its merits in identifying urban change and damage caused by natural disasters such as earthquakes. This paper describes a method for change/damage detection by multiresolution analysis based on wavelet transform. The satellite data set used in here is a pair of Quickbird images acquired before and after the Bam earthquake that occurred on 12/26/2003. For wavelet transform, a family of different mother functions (Daubechies mother wavelet with different orders) have been used and evaluated in this research. Based on the frequency characteristics of the images and the root mean square radius of each scaled mother function, damaged areas can be assessed. After image coregistration phase, mother functions have been evaluated. As a result, each data has been decomposed into five independent images with five different scales (multi-resolution). A window (3 pixels by 3 pixels) has been applied to compensate for miss registration errors on high frequency components in each resulted image. A threshold value has been assigned to identify the pixels with high before-after differences (i.e. changes). In the higher resolution component (scale 1) for the Daubechies of order 1 (Haar function), computational results are impressive and show more than 90% precision as compared with individual counting of the damaged buildings from the image manually. However, this component lacks the visual perception (but statistically more accurate). But, lower resolution (scale 5) components visualize effectively the damage map at proper dimensions with the expense of loosing fine details.