| Techniques in post-disaster assessment from remote sensing imagery have been studied by different research communities in the past decade. Such an assessment benefits everybody from government organizations and insurance agencies to individual home owners. This work explores the application of existing and novel computer vision algorithms for an automated damage assessment caused by windstorm. The various subproblems studied include geometric and photometric correction, rooftop recognition and change classification based on textural differences. Past work done in this area by remote sensing, geoscience, civil engineering and image processing communities had established that the problems addressed in these areas were challenging and largely unsolved. The solutions proposed in this work are strongly motivated towards building a system capable of fast, robust and fine-grained damage analysis from aerial or satellite imagery. The algorithms introduced are thoroughly evaluated and compared with previous works. The results demonstrate that this work promises higher leaps in the field of automated damage classification and provides insights into the reliability of such analysis in real world scenarios. |
