6.1 Introduction
Sea ice information in the polar regions is essential for various applications,
especially for climate change study and marine navigation. Sea ice is an integral
part of the Earth’s climate system. It interacts with both the ocean and the atmosphere by modulating the heat and moisture fluxes. The state and dynamics of sea
ice determines and alters the surface albedo and salt-freshwater redistributions
(IGOS 2007). In particular, because of its high albedo, sea ice tends to reflect
most of the sunlight back to the atmosphere, resulting in cold climate in the polar
regions. However, with the current trend of the contraction of Arctic ice caps
(Kwok et al. 2009), the failure in controlling the amount of sun ray may cause
climate change in local or global scale, leading to serious influence on human life,
the earth’s ecosystem and natural environment. Consequently, sea ice information,
such as extent, type, concentration and thickness, has been recognized as an
Essential Climate Variable by both the World Meteorological Organization
(WMO) and the United Nations Framework Convention on Climate Change
(UNFCCC). Moreover, the sea ice information is essential for ensuring safe marine
navigation. The Northern Sea Route (NSR) in the Arctic region is the shortest
sailing route linking northwestern Europe and northeastern Asia. However, the
navigation in this region is greatly hampered by the presence of sea ice and iceberg
(Johannessen et al. 2006). Therefore, information regarding the conditions and
distributions of different sea ice types is required for reducing hazards of marine
transportation and offshore operations.
While optical sensors can be used for obtaining sea ice information, they depend
on weather condition or sun-light illumination, and are therefore limited by clouds
and darkness. Satellite synthetic aperture radar (SAR), due to its ability to penetrate
the cloud and work day and night, provides a powerful tool for sea-ice monitoring.
RADARSAT-1 and -2 have been the primary source for sea ice monitoring. At the
Canadian Ice Service (CIS), the operational interpretation of SAR sea ice images
relies on human operators to manually process a great number of image scenes
annually. The sea ice charts, as the final product, label each identified region with an
egg code, which indicates sea ice information (e.g., the type, concentration, stage of
development, and floe size). This visual interpretation of SAR sea ice images,
although capable of incorporating human knowledge and experiences, is very
demanding due to the vast amount of daily sea ice observations. Hence, there is
an urgent need to develop automatic programs that are capable of accurately and
time-efficiently discerning the types and extends of different sea ice from SAR
imagery.
This chapter is organized as follows. Section 6.2 describes the principles of SAR
imaging. Section 6.3 summarizes the available satellite SAR sensors. The automatic segmentation of SAR sea ice imagery and the challenges are described in
Sect. 6.4. The proposed method for SAR sea ice segmentation is introduced in
Sect. 6.5. The results of experiments on both simulated and real SAR sea ice images
are presented in Sect. 6.6. Lastly, Sect. 6.7 concludes the study.
114
L. Xu and J. Li
Sea ice information in the polar regions is essential for various applications,
especially for climate change study and marine navigation. Sea ice is an integral
part of the Earth’s climate system. It interacts with both the ocean and the atmosphere by modulating the heat and moisture fluxes. The state and dynamics of sea
ice determines and alters the surface albedo and salt-freshwater redistributions
(IGOS 2007). In particular, because of its high albedo, sea ice tends to reflect
most of the sunlight back to the atmosphere, resulting in cold climate in the polar
regions. However, with the current trend of the contraction of Arctic ice caps
(Kwok et al. 2009), the failure in controlling the amount of sun ray may cause
climate change in local or global scale, leading to serious influence on human life,
the earth’s ecosystem and natural environment. Consequently, sea ice information,
such as extent, type, concentration and thickness, has been recognized as an
Essential Climate Variable by both the World Meteorological Organization
(WMO) and the United Nations Framework Convention on Climate Change
(UNFCCC). Moreover, the sea ice information is essential for ensuring safe marine
navigation. The Northern Sea Route (NSR) in the Arctic region is the shortest
sailing route linking northwestern Europe and northeastern Asia. However, the
navigation in this region is greatly hampered by the presence of sea ice and iceberg
(Johannessen et al. 2006). Therefore, information regarding the conditions and
distributions of different sea ice types is required for reducing hazards of marine
transportation and offshore operations.
While optical sensors can be used for obtaining sea ice information, they depend
on weather condition or sun-light illumination, and are therefore limited by clouds
and darkness. Satellite synthetic aperture radar (SAR), due to its ability to penetrate
the cloud and work day and night, provides a powerful tool for sea-ice monitoring.
RADARSAT-1 and -2 have been the primary source for sea ice monitoring. At the
Canadian Ice Service (CIS), the operational interpretation of SAR sea ice images
relies on human operators to manually process a great number of image scenes
annually. The sea ice charts, as the final product, label each identified region with an
egg code, which indicates sea ice information (e.g., the type, concentration, stage of
development, and floe size). This visual interpretation of SAR sea ice images,
although capable of incorporating human knowledge and experiences, is very
demanding due to the vast amount of daily sea ice observations. Hence, there is
an urgent need to develop automatic programs that are capable of accurately and
time-efficiently discerning the types and extends of different sea ice from SAR
imagery.
This chapter is organized as follows. Section 6.2 describes the principles of SAR
imaging. Section 6.3 summarizes the available satellite SAR sensors. The automatic segmentation of SAR sea ice imagery and the challenges are described in
Sect. 6.4. The proposed method for SAR sea ice segmentation is introduced in
Sect. 6.5. The results of experiments on both simulated and real SAR sea ice images
are presented in Sect. 6.6. Lastly, Sect. 6.7 concludes the study.
114
L. Xu and J. Li