In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Voi. XXXVIII, Part 7B 
AUTOMATIC 3D CHANGE DETECTION BASED ON OPTICAL SATELLITE STEREO 
IMAGERY 
J. Tian, H. Chaabouni-Chouayakh, P. Reinartz, T. Krauß, P. d’Angelo 
German Aerospace Center (DLR), Remote Sensing Technology Institute, 82234 Wessling, Germany - 
(Jiaojiao.Tian, Houda.Chaabouni, Peter.Reinartz, Thomas.Krauss, Pablo.Angelo)@dlr.de 
KEY WORDS: Optical Stereo Data, DSM, Change Detection, Building, 3D-Analysis 
ABSTRACT: 
When monitoring urban areas from space, change detection based on satellite images is one of the most heavily investigated topics. 
In the case of monitoring change in 2D, one major shortcoming consists in the lack of height change detection. Thereby only changes 
related to reflectance values or local textures changes can be detected. However, changes in the vertical direction are completely 
ignored. In this paper we present a new 3D change detection approach. We focus our work on the detection of changes using Digital 
Surface Models (DSMs) which are generated from stereo imagery acquired at two different epochs. The so called “difference image” 
method is adopted in this framework where the final DSM is subtracted from the initial one to get the height difference. Our 
approach is a two-step approach. While in the first step, reduction of the noise effects (coming from registration noise, matching 
artifacts caused by the DEM generation procedures, etc), the second one exploits the rectangular property of the building shape in 
order to provide an accurate urban area monitoring change map. The method is tested, evaluated and compared with manually 
extraction results over the city centre of Munich in Germany 
1. INTRODUCTION 
Change detection using automated image processing methods 
is a very important topic in satellite image processing. 
Numerous detection methods using various image types have 
been developed to satisfy a wide range of applications and 
user requirements (e.g. Singh 1989, Bruzzone 1997, Lu 
2004). One major problem often met when restricting the 
change detection to the 2D information extracted from 
satellite images, is the lack of monitoring height changes, the 
3D component of the surface to be analyzed. Thereby only 
changes related to the reflectance values and/or local textural 
changes are detected. However, changes in the vertical 
direction, such as building height changes are completely 
ignored. Such information could play an important role in 
different applications such as disaster assessment and urban 
area construction and/or destruction monitoring. Moreover, 
with the increasing availability of high resolution stereo 
imagery acquisition as well as the steady development of 
automatic DSM generation techniques (Zhang, 2005; Krauß, 
2007; Akca, 2007; d'Angelo, 2008), comparison of the higher 
resolution DSMs acquired at different epochs on a same 
urban area should provide valuable information about the 
potential changes that have occurred at higher levels (e.g. 
building construction/destruction). 
In the literature, several studies have been dedicated to the 
detection of changes using DSMs generated from stereo 
imagery. They can mainly be divided into two categories. 
The first change detection method is based on the joint-use of 
stereo and multi-spectral images (when they are available) 
and the generated DSMs are used in order to detect the 
changes that occur in the 2D space (spatial changes) as well 
as in the 3D space (height changes). In fact, the DSMs do not 
include spectral or textural information which could be of 
great help when the task is to perform an accurate change 
monitoring. For example, Sasagawa (Sasagawa, 2008) 
integrated the DSM-difference map with the multispectral 
satellite images as an input to manual interpretation. Krauß 
(2007) applied a vegetation mask derived from multispectral 
data to the DSMs in order to concentrate only on urban 
structure changes. The second change detection method is 
based on DSM difference (when stereo or multispectral data 
is not available like in the case of Laser DSMs). The changed 
areas are detected through a simple subtraction of one DSM 
from another. This approach has been used in several 
researches (Zhang, 2005; Reinartz, 2006; Akca, 2007) for 
DSM precision assessment tasks. However, in this category 
of 3D change detection, the quality of the generated DSMs is 
quite determining regarding the accuracy of the final change 
maps. In fact, miss-coregistrations and significant height 
differences that may arise between DSMs generated from 
different sources often results in the detection of virtual or 
irrelevant changes. In the work of Chaabouni-Chouayakh 
(2010) for example, post-processing steps such as 
morphological operations and contextual knowledge 
introduction have been proposed to remove virtual changes 
and to keep only the real ones. 
In this paper, we focus on the detection of urban area changes 
(building construction/destruction). Our work includes 
vertical change as well as horizontal change. In this paper, 
the vertical change means the changes in height direction, 
while horizontal changes mentions the changes in planimetrie 
direction, especially to detect the changing size of the subject 
in x andy. For the vertical changes, we compute the so-called 
“difference image” (Singh, 1989; Fung, 1990, Bruzzone, 
2000) between two DSMs acquired over the city centre of 
Munich between 2003 and 2005. We adopt the robust image 
differencing method to eliminate the noise edges. After that,