PROJECT TITLE :

Estimation of Seismic Vulnerability Levels of Urban Structures With Multisensor Remote Sensing

ABSTRACT:

The ongoing global transformation of human habitats from rural villages to ever growing urban agglomerations induces unprecedented seismic risks in earthquake prone regions. To mitigate affiliated perils needs the seismic assessment of built environments. Various studies emphasize that remote sensing can play a valuable role in supporting the extraction of relevant features for preevent vulnerability analysis. However, the bulk of approaches operate on building level. This induces the deployment of terribly high spatial resolution remote sensing information, which hampers, these days, utilization capabilities for larger areas because of data prices and processing necessities. In this paper, we have a tendency to alter the spatial scale of study and propose ideas and methods to estimate the seismic vulnerability level of homogeneous urban structures. A procedure is designed, that includes four main steps dedicated to: one) delineation of urban structures by suggests that of a tailored unsupervised data segmentation procedure with scale optimization; a pair of) characterization of urban structures by a joint exploitation of multisensor knowledge; three) choice of most possible features below consideration of in situ vulnerability info; and 4) estimation of seismic vulnerability levels of urban structures within a supervised learning framework. We render the prediction problem in 3 ways that to handle operational requirements that can evolve in real-life things. one) To discriminate 2 or more categories based mostly on labeled samples of all categories gift in the info beneath investigation, we use the framework of soft margin support vector machines (C-SVM). 2) To think about situations, where solely labeled samples are on the market for the class(es) of interest and not for all categories present in the data, we deploy ensembles of $nu$-one-category SVM ($nu $-OC-SVM). and three) To work knowledge with the next statistical level of measurement (interval or ratio scale), we tend to utilize a support vector regression (SVR) approach to estimate a regression function from the training samples. Experimental results are obtained for the earthquake-prone mega city Istanbul, Turkey. We have a tendency to use multispectral data from the RapidEye constellation, elevation measurements from the TanDEM-X mission, and spatiotemporal analyses based mostly on information from the Landsat archive to characterize the urban environment. Moreover, different in situ data sets are incorporated for Istanbul’s district Zeytinburnu and therefore the residual settlement area of Istanbul. When estimating injury grades for Zeytinburnu with SVR, best models are characterized by mean absolute share errors less than 11p.c, and fairly strong goodness of match ($R > 0.75$). When aiming to identify completely different types of urban structures for the remaining settlement space of Istanbul (i.e., urban structures determined by giant industrial/industrial buildings and tall detached residential buildings, that can be thought of here as highly and slightly vulnerable, respectively), results obtained with $C$-SVM show a particular increase of accuracy compared to results obtained with ensembles of $nu$-OC-SVM. The latter were not able to exceed moderate agreements, with $kappa$ statistics slightly on top of zero.forty five. Instead, $C$-SVM allowed getting $kappa $ statistics expressing sub