Following a natural disaster, a comprehensive evaluation of the damage caused is the prerequisite for any response and rehabilitation activities. Depending on disaster type and severity, however, such assessment is impeded by (i) the inhomogeneous nature of damage, (ii) potentially wide-spread consequences, (iii) the possibility of delayed secondary damage, and (iv) lack of information from, and restricted access to the disaster site.

The spatial and temporal characteristics of a disaster create a potential for remote sensing to aid in damage assessment, although they must match the properties of the sensor used. While spaced-based platforms eliminate the need for aircraft deployment to the disaster site (which may not be possible for logistical or meteorological reasons), the temporal, and frequently also spatial resolution is still insufficient. Especially for urban damage assessment, even recent high-resolution sensors such as Ikonos and QuickBird show limitations in resolving structural elements at the required scale. In fact, the very nature of structural damage reduces the utility of vertically acquired data, also those obtained from aerial platforms. While researchers have had some success identifying urban disaster damage, this has mostly been true for areas that have suffered near-complete destruction, for example collapse following an earthquake. The reason for this is that only total collapse of buildings leads to the spectral changes that can be identified in texture analysis, or results in an identifiable narrowing of streets due to extensive building rubble.

The problem with vertically acquired image data is that the building roofs, the only readily accessible structural element, are poor indicators of the damage actually sustained by a building. Frequently, compromised structural integrity is expressed by buildings leaning or slumping, and by façade and column cracks, while the roof is left largely intact. Laser scanning has been used to assess changes in building heights, although such change detection requires rarely available pre-event datasets.

Because of the limitations of vertical imagery, several studies have investigated the potential of oblique and transverse images. Single photographs taken from helicopters have been subjected to texture analysis to identify damage. In a more sophisticated approach, close-range photogrammetry has been used to map damage to individual façades and houses in detail.

The purpose of this project is to investigate the possibility of (i) a 3D photogrammetric reconstruction of an urban disaster site, (ii) the effective linking of the results to a GIS database, and (iii) the optimal integration of different image data and sensing types. Following a disaster in urban areas, TV footage tends to be the first available data. Photogrammetric principles can be applied to individual frames extracted from a video stream, with the aim of reconstructing the affected site at a level of detail that allows façade damage to be detected. The focus of the research would be on working out the acquisition of required ground control, and the processing of data from an uncalibrated camera. Questions related to obstructed views in densely built-up areas also need to be addressed.

Part (ii) of the study will investigate how identified damage can be linked to an existing GIS database, with the aim of inventorising damage by building type, but also to update the database. Lastly, it is important to realise that data available shortly after an event are rarely perfect, and that the most suitable, but unavailable data types have to be replaced by alternatives, and that best results may be achieved by integrating different data. In this context, the utility of images with varying resolution and quality characteristics could also be investigated. Further, depending on data availability, the integration of laser scanning or hyperspectral data would be useful.

Interested candidates can contact Dr. Norman Kerle (e-mail: kerle@itc.nl) for more information.