FUSION SAR AND OPTICAL IMAGES TO DETECT OBJECT-SPECIFIC CHANGES 
Mu H. Wang ab , Hai T. Li a , Ji.X Zhang 3 ,Jing H.Yang 3 
a) Chinese Academy of Surveying and Mapping,Beijing,l 00039,China-(lhtao,zhangjx,)@casm.ac.cn 
b) Liaoning Technical University ,School of Geomaticas,Fuxin, 123000,China- amudc@126.com 
Commission VII Working Group VII/6 
KEYWORDS: SAR, Optical images, Fusion, Segmentation, Change Detection, Object-Specific 
ABSTRACT: 
Although SAR images have been widely used in the field of change detection because of the capability of SAR sensors to acquire 
images during day and night as well as all weather conditions, the information provided by SAR data alone may not sufficient for 
detailed object analysis.We explore an approach based on fusion of SAR and optical images to detect object-specific changes.The 
assumption of the proposed method is that we can acquire SAR data and corresponding optical data respectively at different 
periods.The approach in this paper consists of three steps:images pre-processing, SAR and optical images fusion, change detection 
based on region features.Specifically speaking, geometric or radiometric calibration and precise registration between SAR and 
optical images are included in the first phase. In the second step,we adopt wavelet transform-based fusion method.Multiscale image 
segmentation approach is separately applied to the two fused image for obtaining object region. Then we compare the corresponding 
region features extracted from the two segmented images using distance distinguishing function. The experiment results illustrate that 
the approach can make use of the complemental characteristics of SAR and optical images to acquire more information and 
effectively detect object changes. 
1. INTRODUCTION 
Remote sensing technique has been widely used in the field of 
change detection because of the advantage of macroscopy, high 
speed and short interval of acquiring resources, ample 
information and effective usability (Massonnet et al., 
1993).While Synthetic Aperture Radar (SAR) has the capability 
to obtain information during day and night as well as all 
weather conditions, which supplies a gap of optical images. The 
information extracting from fusion of SAR and optical images 
is complementary for detailed object analysis. 
In recent years, many change detection techniques have been 
developed. They can broadly group into three categories (Baud 
ouinet ah, 2006): visual interpretation, pixel-based approaches, 
and object-based methods. Specifically, visual interpretation 
requires human experience (computer-assisted or not) to label 
zones that are considered as changed, which can make full use 
of analysts’ experience and knowledge but is time-consuming. 
Digital pixel-based approaches compare the spectral features 
between pixels on low or middle resolution remote sensing 
images. There are many kinds of methods in this category, 
which performances are rarely compared to each other because 
these techniques are considered scene-dependent(Lyon et 
ah, 1998, Rogan et ah,2003).The main drawback of pixel-based 
methods are the “salt and pepper” effect in the resulting map. 
Moreover, the spatial or contextual information between 
proximate pixels is most often ignored(Atkinson et al.,2000, 
Townshend et ah,2000).Object-based methods combine the 
advantage of visual interpretation and pixel-based approaches 
and incorporate spatial, texture and structure information of 
pixel groups, which are spatially continuous and homogeneous 
regions(named objects) dividing by image segmentation 
techniques on satellite images. Object-based methods make full 
use of contextural information (Flanders et ah,2003) between 
super- and sub-objects as well as neighbor objects. However, 
many of these methods suffered of low detection performance. 
This research aims to develop a new region-based method to 
detect symbolic buildings’ changes collapsed during the Bam 
earthquake, taking advantage of multi-resolution 
segmentation Baatz et ah, 1999),region features extraction, 
associated regions search and features similarity comparison. 
Furthermore, the approach proposed in this paper is tested on a 
multi-temporal fusion SAR and optical images and compared its 
performance to visual interpretation method and a pixel-based 
method in ENVI 4.3. 
2. STUDY SITE AND DATA 
The study site is located in Bam,Iran and the epicenter suitates 
at 29.002°N, 58.325° E.In the test we adopt ENVISAT ASAR 
data and multi-spectral QuickBird satellite images. Table 1 lists 
the information of high-resolution QuickBird imagery of Bam. 
Figure 1 shows the time sequence and track number of seven 
breadthes of ASAR data.Considering the epicenter location and 
happen time of Bam earthquake,we choose two breadthes which 
track numbers are 9192 and 9693. 
Acquisition Date 
Spatial Resolution 
Image Size 
Image pre-processing