OBJECT-BASED CHANGE DETECTION
USING HIGH-RESOLUTION REMOTELY SENSED DATA AND GIS
N. Sofina
a
, M. Ehlers
b
a
University of Osnabrueck, Institute for Geoinformatics and Remote Sensing, 49076 Osnabrueck, Barbarastrasse 22b,
Germany - nsofina@igf.uni-osnabrueck.de
b
University of Osnabrueck, Institute for Geoinformatics and Remote Sensing, 49076 Osnabrueck, Barbarastrasse 22b,
Germany - manfred.ehlers@uos.de
Commission VII, WG VII/5
KEY WORDS: Change Detection, Geographic Information Systems (GIS), Detected Part of Contour, Generation of Features, Data
Mining, GIS GRASS, Python
ABSTRACT:
High resolution remotely sensed images provide current, detailed, and accurate information for large areas of the earth surface which
can be used for change detection analyses. Conventional methods of image processing permit detection of changes by comparing
remotely sensed multitemporal images. However, for performing a successful analysis it is desirable to take images from the same
sensor which should be acquired at the same time of season, at the same time of a day, and – for electro-optical sensors - in cloudless
conditions. Thus, a change detection analysis could be problematic especially for sudden catastrophic events. A promising alternative
is the use of vector-based maps containing information about the original urban layout which can be related to a single image
obtained after the catastrophe. The paper describes a methodology for an object-based search of destroyed buildings as a
consequence of a natural or man-made catastrophe (e.g., earthquakes, flooding, civil war). The analysis is based on remotely sensed
and vector GIS data. It includes three main steps: (i) generation of features describing the state of buildings; (ii) classification of
building conditions; and (iii) data import into a GIS. One of the proposed features is a newly developed ‘Detected Part of Contour’
(DPC). Additionally, several features based on the analysis of textural information corresponding to the investigated vector objects
are calculated. The method is applied to remotely sensed images of areas that have been subjected to an earthquake. The results show
the high reliability of the DPC feature as an indicator for change.
1. INTRODUCTION
Since the 20-th century the integration of cartography,
Geoinformatics, and remote sensing technologies has been
developed due to rapid technological advances. This evolution
implied a number of derivative scientific directions and
consolidation of conventional and digital cartographic
techniques.
The technological progress in remote sensing was one of the
most important origins of this integration. Its influence on the
cartography is many-folded. The main aspect is that remotely
sensed images are considered as maps which lead to the
development of new photogrammetry and image processing
methods.
A second origin of the integration tendencies was the
development of geoinformation technologies. In recent years the
application of GIS has offered the greatest potential for
processing large volumes of multi-source data (Ehlers et al.,
1989).
For change detection of the Earth’s surface the development of
methods based on integrated analysis of vector and image data
is a point of general interest. The majority of change detection
methods address land-use change dynamics (see, for example,
Centeno & Jorge, 2000; Lo & Shipman, 1990; Li, 2009;
Mattikalli, 1995; Weng, 2002) and only a few deal with damage
assessment. Samadzadegan & Rastiveisi (2008) used
information from pre-event vector maps and post-event satellite
images to compare different textural features for extracted
building vector. Chesnel et al. (2007) used a pair of very high
resolution images of a crisis region together with GIS data for
investigation based on different acquisition angles of the
images.
This paper presents a GIS-based approach for the detection of
destroyed buildings which were affected by a catastrophic event
like an earthquake, a landslide or a military action. The
methodology is based on the integrated analysis of vector data
containing information about the original urban layout and
remotely sensed images obtained after a catastrophic event. The
proposed method was applied to damage mapping after the
earthquake in Qinghai, China (2010). GIS information was
obtained by digitizing from pre-earthquake images. The
performed experiments indicate that a GIS-based analysis for
change detection can essentially improve the potential of
remotely sensed data interpretation and can be considered as a
powerful tool for the detection of destroyed building.
The proposed methodology was developed solely within the
Open Source software environment (GRASS GIS, Python,
Orange). The use of Open Source software provides an
innovative, flexible and cost-effective solution for change
detection analyses.
2. STUDY AREA
To assess the effectiveness of the proposed approach, the
experiments were conducted using a high-resolution
panchromatic QuickBird image acquired after the earthquake
that struck Yushu, Qinghai, China on April 14, 2010 with a
magnitude of 6.9Mw. GIS information was obtained by
digitizing from pre-earthquake images.
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012
XXII ISPRS Congress, 25 August – 01 September 2012, Melbourne, Australia