1077
i
Beijing 2008
CHANGE DETECTION APPROACH TO SAR AND OPTICAL IMAGE INTEGRATION
Yu ZENG a ’\ Jixian ZHANG b , J.L.VAN GENDEREN c
Shandong University of Science and Technology, Qingdao, Shandong Province, 266510, P.R.China
b Chinese Academy of Surveying and Mapping, Beijing, 100039, P.R.China, zengyu@casm.ac.cn
c International Institute for Aerospace Survey and Earth Sciences (ITC),
P.O. Box 6, 7500 AA Enschede, The Netherlands
Commission VII, WG VII/6
KEY WORDS: Environmental Monitoring, Change Detection, Fusion, Land Use, Land Cover, SPOT5, RADARSAT
ABSTRACT:
In order to overcome the insufficiency of single remote sensing data source during information extraction, to make use of the
complementary characteristics of SAR data and optical imagery, and to facilitate better monitoring and evaluation of resources and
ecological environment, this paper develops the idea and presents the approach to land use/cover change detection by different
temporal SAR and optical image integration. Aiming at the different imaging mechanisms and information characteristics of optical
imagery and SAR data, this paper employs the object-oriented image analysis technique for high accuracy information extraction
from high-resolution optical imagery; it proposes a multi-scale and multi-texture feature fusion method based on SVM for
information extraction from single-band and single-polarization SAR data by employing the multi-scale textural analysis technique
and the fractal analysis technique; it then integrates individually extracted information of different temporal for change information
extraction by applying a series of decision rules; it also develops the method for analysis and evaluation of uncertainty of the change
detection result at the scale of pixels using the extended probability vector. Data adopted in this research are different temporal
SPOT5 image and Radarsat-1 SAR data. Experimental results prove the correctness, reliability and effectiveness of the methods
proposed in this paper.
1. INTRODUCTION
Timely and accurate change detection of Earth’s surface
features is extremely important for understanding relationships
and interactions between human and natural phenomena in
order to effectively manage and use resources as well as to
promote better decision making. Because of the advantages of
repetitive data acquisition, its synoptic view, and digital format
suitable for computer processing, remote sensing data, such as
AVHRR, Landsat TM/ETM + , SPOT, IKONOS, SAR and aerial
photographs, have become the major data sources for different
change detection applications during the past decades.
Optical sensors acquire the reflected energy from sunlight
reaching the ground in the visible and near-infrared spectrum.
While the multi-spectral data presents the rich spectral
information of the observed objects, the panchromatic data
which is often available with a higher resolution than multi-
spectral data shows detailed geometric information of the
objects. These features make optical imagery relatively easier
for interpretation and become the main remote sensing data
source for change detection. However, due to the limitation of
data acquisition, such as the impact of clouds, fog, or smoke on
optical sensors, sometimes, it is difficult to obtain the same
sensor optical images that meet the temporal requirement.
Different from optical sensors, a SAR sensor has the all-
weather capability. Factors which influence the intensity of
radar returns determine that SAR data can present rich
structural and texture information, and is sensitive to water
body, vegetation, and built-up areas. However, due to the
limited band numbers and polarization modes, as well as the
affects caused by speckle noises, slant-range imaging,
foreshortening, layover, and shadows, SAR data is relatively
difficult for interpretation compared with the optical imagery.
In order to overcome the insufficiency of single remote sensing
data source during information extraction, to make use of the
complementary characteristics of SAR data and optical imagery,
and to facilitate better monitoring and evaluation of resources
and ecological environment, this paper develops the idea and
presents the approach to land use/cover change detection by
different temporal SAR and optical image integration. Test site
of this research is located at Jinnan district, Tianjin, China. Data
adopted are SPOT5 Pan/XS image acquired on October 16,
2004 and SGF F3 Radarsat-1 SAR data acquired on October 19,
2005 with the spatial resolution 6.25m. The rest of the paper is
organized as follows. Section 2 gives an overview on the
methodology of this research. Section 3 describes the
experimental results and their analysis and conclusions are
drawn in Section 4.
2. METHODS
2.1 Object-oriented Image Analysis Technique
Different from pixel-based image classification approach,
object-oriented image classification operates on objects instead
of single pixels. These objects consist of many pixels that have
been grouped together in a certain way by image segmentation,
and each object is homogeneous and non-intersecting with the
others. Object-oriented image analysis technique uses shape,
size, textural, contextual information as well as spectral
information to perform information extraction at the level of
objects. While high spatial resolution remote sensing provides