Severe hurricane events have been occurring across the United States, threatening both highly developed urban areas and distressed island communities. Assessment of building damages due to hurricane events is a critical element in disaster management as it supports not only search and rescue operations but also provides insights into the performance of existing planning and building practices. But unlike many other extreme weather events, hurricanes can topple an entire region for an extended amount of time, creating a daunting task for traditional foot-on-ground building damage assessment approaches.

The overarching goal of this study is to create and test a computational framework to leverage big spatial data acquisition and processing technologies for automated building damage assessment. The specific aims of this study include: (1) formulating a cohesive and multi-scale damage assessment approach that considers the continuously evolving data sources and damage assessment needs during different phases of disaster management; (2) developing algorithms for rapid building damage assessment with airborne lidar data, which are typically collected immediately after the landing of hurricane events; (2) developing algorithms for component-level building damage assessment with high-resolution ground-based lidar data; (3) charactering the performance of image based 3D reconstruction for damage assessment; (4) developing robust image alignment algorithms for geo-registering post-disaster image data from varied sources to realize the fusion of heterogeneous point cloud and image data for comprehensive damage assessment. The proposed methods were applied on several geospatial data sets collected during Hurricane Sandy. The results are compared with the ground truth which was created by a manual labeling process. The results show that the proposed methods are capable of conducting damage assessment of building structures autonomously and at different resolution and extracting useful damage information to support building performance modeling. Future research of this study will be focused on leveraging high performance computing capabilities to accelerate the damage assessment process.