In: Wagner W„ Szäkely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B
644
CHANGE ANALYSIS WITH TERRASAR-X DATA
D. Weihing 3 ’ *, F. von Poncet 3 , M. Schlund b , O. Lang 3
3 Infoterra GmbH, Claude-Domier-Strasse, 88090 Immenstaad, Germany - (Diana.Weihing, Felicitas.Poncet,
01iver.Lang)@infoterra-global.com
b Department of Geography, University of Jena, Germany - Michael.Schlund@uni-jena.de
Commission VII
KEY WORDS: Analysis, Change Detection, Monitoring, Change, SAR
ABSTRACT:
Change analyses play an important role for different applications, ranging from small- to large-scale monitoring. To identify changes
from SAR images of repeat pass acquisitions different methods are commonly applied, which differ with respect to the parameter as
an indicator of changes. The backscatter intensity is analysed in incoherent change detection methods, whereas in the coherent
change detection the complex correlation coefficient is analysed as a change indicator. These methods provide complementary
characterisations, since they are sensitive to different measures of a SAR scene.
However, often not only the detection of changes, but also the assessment of the detected changes is of interest, which automated is a
challenging task.
In this paper a combination of detection and assessment of changes is shown. Additional information is integrated into the proposed
scheme to separate relevant changes from less relevant ones in order to decrease unrequested changes and evaluate the type of
change. Texture measures provide information on the spatial variation of the backscatter and thus information of the local surface
characteristics within the scene. This information is used to restrict the search for areas and changes of interest as well as for the
assessment of changes.
The goal of our analysis is to support the implementation of an operational change detection process and to define a suitable
assessment of changes from TerraSAR-X data regarding different customers’ requirements.
1. INTRODUCTION
In many different geoinformation applications, the detection
and assessment of changes is of much interest. These
applications can vary from specific site monitoring to large-
scale assessment after certain events and for updating existing
geoinformation databases.
Satellite SAR missions like TerraSAR-X entail potential for
such applications, not only because of the large coverage in
combination with high resolution, but also due to the
independence regarding daytime and weather conditions.
The identification of changes implies a comparison of datasets
or knowledge about a scene of different dates, like e.g. updating
of existing geoinformation with recent remote sensing data.
Different approaches for change detection have been studied
depending on the sensor and available data.
In this paper the focus is set on the image-to-image change
detection with TerraSAR-X data, and in particular on repeat-
pass images with the same acquisition parameters, which allow
a direct comparison. Therefore, acquisitions of pre- and post
events are required.
Detecting changes means that every variation causing a
different echo signal/backscatter is detected. However, often
customers are only interested in relevant changes concerning
their applications. To be able to offer the user-specific change
information, an assessment of the detections is required. The
preferably automatic detection and assessment is a goal for a
worldwide operational process and service. The integration of
additional information, like e.g. GIS data, in such a process can
be helpful for this purpose. However, in many regions often this
required a-priori information is missing or not available in the
time frame for rapid assessment. Therefore, the used pre- and
post-event imagery for change detection as a single source has
to be exploited also to asses the changes. Hence, different
measures have to be educed from the images to derive more
information about the scene.
2. CHANGE DETECTION
Change detection based on SAR images has been subject in
different publications. To identify changes from SAR images of
repeat pass acquisitions different methods are commonly
applied. These methods differ with respect to the parameter
which is used to indicate changes and the method to reduce the
noise. Since SAR data contains amplitude and phase
information, both parameters can be used as change indicators.
A common method is the analysis of the radar backscatter
intensity in time to identify changes between different
acquisitions, also called incoherent change detection (Preiss et
al., 2003). This power change estimate is affected by the
speckle noise component, whose reduction is discussed in
several publications, e.g. in (Schmitt et ah, 2009).
Corresponding author.