ADVANCED STRUCTURAL DISASTER DAMAGE ASSESSMENT BASED ON AERIAL 
OBLIQUE VIDEO IMAGERY AND INTEGRATED AUXILIARY DATA SOURCES 
N. Kerle? , R. Stekelenburg" 
* Dept. of Earth Observation Sciences, International Institute of Geo-information Science and Earth Observation, 
Hengelosestraat 99, P.O.Box 6, 7500 AA Enschede/The Netherlands (kerle(@ite.nl) 
^ InnoStack, Ommerbos 51, 7543 GG Enschede/The Netherlands (mail innostack.com) 
Commission VII, TS-PS: WG VIS 
KEY WORDS: Disaster, video, satellite, change detection, integration, texture 
ABSTRACT: 
Following natural and man-made disasters, comprehensive and reliable information on the nature, extent, and actual consequences of 
an event is required. Obtaining such information is particularly challenging following unheralded disasters, such as earthquakes or 
industrial accidents. In those situations, currently operational space-based sensors may not be able to provide timely data. In 
addition, even high spatial resolution satellites (« 1m) are limited in their capacity to reveal true 3D structural damage at a level of 
detail necessary for appropriate disaster response. 
In this study we investigated the use of oblique aerial video imagery for systematic quantitative damage assessment. 
Specifically, the following issues were addressed: (i) extraction of individual frames from aerial TV video data, and subsequent 
enhancement of frame information content using a synthetic aperture approach, and by stacking adjacent frames; (ii) spatial 
registration of individual frames based on automatically extracted positional information, and (iii) damage analysis based on HIS 
values and edge elements. We created a working environment that facilitates the video-based damage assessment process and 
integration with auxiliary data. The limited success of the automatic damage assessment was caused by poor image quality and 
empirically determined damage threshold values. We provide recommendations for improved image acquisition, and plan further 
work focusing on generic texture parameters and object geometry. 
As an illustration, we used aerial video data acquired in May 2000 over Enschede, the Netherlands, following the explosion of a 
  
fireworks factory that severely damaged or destroyed nearly 500 buildings and caused 22 fatalities. 
1. INTRODUCTION 
1.1 Motivation 
Natural and man-made disasters create a need for rapid, 
comprehensive and reliable information on the nature, extent, 
and actual consequences of an event. The overall cost of a 
disaster, both in terms of economic damage and fatalities, depends 
on how quickly the event is responded to, and how efficiently 
response activities are managed. Particularly stringent 
information collection constraints are present following 
unheralded disasters, such as earthquakes or industrial 
accidents. The response to virtually every extensive disaster in 
recent years was delayed by a slow inventory of the event’s 
consequences, in developing countries (e.g. following 
earthquakes in Koceali [Turkey; 1999], Gujarat [India; 2001], 
or Bam [Iran; 2003]) as well as developed ones (e.g. following 
the 1995 Kobe, Japan, earthquake). This is due to a shortage of 
reliable information coming from the disaster site, difficulties in 
access, and organizational reasons, such as insufficient 
preparedness. The utility of geoinformatics in general, and 
remote sensing in particular, to provide timely and 
comprehensive information of the post-disaster situation has 
been repeatedly identified (e.g. Alexander, 1991; Walter, 1994). 
The reasoning is that a comprehensive damage inventory, 
prepared at the appropriate (i.e. disaster type-specific) synoptic 
and detailed scales, provides the prerequisite to direct limited 
disaster response resources with maximum benefit. 
A number of sensors are of potential use to acquire the 
necessary data, optical spaceborne ones being the first choice 
due to ease of image interpretation (for example as opposed to 
radar or laser scanner data), and data acquisition and 
distribution infrastructure being unaffected by damage on the 
ground. However, while the potential synoptic coverage of 
spaceborne systems is certainly an asset, in addition to possible 
data capture delays due to orbit and pointability restrictions 
(Kerle and Oppenheimer, 2002), currently operational sensors 
may not be able to provide a detailed damage assessment in 
urban areas. Even high spatial resolution sensors (< Im) are 
limited in their capacity to reveal true 3D structural damage at a 
level of detail necessary for appropriate disaster response. This 
is because the vertical view largely restricts information to 
building roofs, which may remain intact despite extensive 
structural damage more readily expressed on building façades. 
The same is also true for standard aerial photography, which 
provides the best data in terms of spatial resolution, but which, 
while typically not being available shortly after a disaster, also 
tends to suffer from the limitation of vertical viewing. 
1.2 The potential of aerial video data 
The first data type likely to be available after a disaster in an 
urban setting is oblique airborne video imagery captured by the 
news media. The imagery acquired, however, differs 
substantially from standard air- and spaceborne remote sensing, 
580 
Interna. 
both in 
surveys 
scheme 
comple 
compar 
used ai 
lines, rt 
1920 c 
additioi 
increas: 
disaster 
infrarec 
While t 
such d: 
method 
include 
platfort 
image : 
lack cai 
The 19 
data fi 
overpas 
days; J 
work b 
interpre 
data o 
Hasega 
applied 
showin 
HDTV 
were a] 
and Tu 
disaster 
data to 
were a 
practic 
1.3 Ri 
A num 
limited 
compo 
conver: 
quality 
method 
synthet 
Gornyi 
assessn 
frames 
used ir 
azimut! 
inform: 
flightp: 
The ov 
video 
elemen 
data to 
geocod 
An imr 
With ar 
quite d 
to our