92
S.G. BEAVEN AND S.P. GOGINENI
developed a technique for incorporating information from the ERS-1 SAR into a passive microwave algorithm to improve ice type concentration estimates. We used ERS-1
SAR data to estimate multiyear ice fraction through a dynamic thresholding technique.
Using the SAR-derived multiyear ice concentration in a modified version of the NT algorithm we obtained an improved first-year ice concentration estimate. This technique
demonstrates that the use of both active and passive satellite data can remove or reduce
the ambiguities that occur in estimating ice type concentration.
5.2
Data Fusion Background
5.2.1
Data Fusion for Sea Ice Remote Sensing
It has been speculated that a combination of active and passive microwave measurements will be important in sea ice remote sensing (Cavalieri et a1.1990). A number of
multisensor studies of the microwave properties of sea ice have been carried out over
the last three decades with the primary goal of determining backscatter and emission
characteristics of sea ice. Current techniques for exploitation of satellite microwave data
use only single sensor information. The use of single-sensor data results in ambiguities in the estimations of geophysical parameters. The fusion of multisensor data has
been discussed as a means for removing these ambiguities. Current spaceborne SAR
systems used by the remote sensing community (ERS-1 and RADARSAT) are capable
of resolution of a few tens of meters, whereas the current series of spaceborne passive
microwave imagers (SSM!I) have spatial resolution on the order of tens of kilometers,
depending on the wavelength. This limits techniques for combining these data sets to
those that can overcome or exploit this disparity in resolution.
Combining active and passive imagery has been previously suggested as an approach
to improving estimates of key sea ice geophysical parameters such as thickness, mass
distribution and extent (Burns et a1.1987; Livingstone, et a1.1987; Drinkwater et a1.1991;
Collins 1992; Grenfell et a1.1992). However, until recently multisensor fusion approaches have not been applied to this problem. Below we discuss the three basic architectures
for multisensor fusion, followed by a hybrid approach that shows promise for fusion of
active and passive microwave data.
5.2.2
Data Fusion Architectures
The purpose of fusing multisensor data is to obtain better estimates of geophysical parameters or new information that could not be obtained with any single sensor. Three
basic architectures are utilized in multisensor data fusion processes. These are characterized by the level at which the data are combined or fused - data level, feature level
and decision level (Waltz and Llinas 1990) - and are defined for an identity declaration
process requiring fusion of multisensor data. Although multisensor fusion problems
may not necessarily fall into the identity declaration category, any fusion process is typically characterized by one of these three categories, or a hybrid of these. The architectures are shown in Fig.!.
S.G. BEAVEN AND S.P. GOGINENI
developed a technique for incorporating information from the ERS-1 SAR into a passive microwave algorithm to improve ice type concentration estimates. We used ERS-1
SAR data to estimate multiyear ice fraction through a dynamic thresholding technique.
Using the SAR-derived multiyear ice concentration in a modified version of the NT algorithm we obtained an improved first-year ice concentration estimate. This technique
demonstrates that the use of both active and passive satellite data can remove or reduce
the ambiguities that occur in estimating ice type concentration.
5.2
Data Fusion Background
5.2.1
Data Fusion for Sea Ice Remote Sensing
It has been speculated that a combination of active and passive microwave measurements will be important in sea ice remote sensing (Cavalieri et a1.1990). A number of
multisensor studies of the microwave properties of sea ice have been carried out over
the last three decades with the primary goal of determining backscatter and emission
characteristics of sea ice. Current techniques for exploitation of satellite microwave data
use only single sensor information. The use of single-sensor data results in ambiguities in the estimations of geophysical parameters. The fusion of multisensor data has
been discussed as a means for removing these ambiguities. Current spaceborne SAR
systems used by the remote sensing community (ERS-1 and RADARSAT) are capable
of resolution of a few tens of meters, whereas the current series of spaceborne passive
microwave imagers (SSM!I) have spatial resolution on the order of tens of kilometers,
depending on the wavelength. This limits techniques for combining these data sets to
those that can overcome or exploit this disparity in resolution.
Combining active and passive imagery has been previously suggested as an approach
to improving estimates of key sea ice geophysical parameters such as thickness, mass
distribution and extent (Burns et a1.1987; Livingstone, et a1.1987; Drinkwater et a1.1991;
Collins 1992; Grenfell et a1.1992). However, until recently multisensor fusion approaches have not been applied to this problem. Below we discuss the three basic architectures
for multisensor fusion, followed by a hybrid approach that shows promise for fusion of
active and passive microwave data.
5.2.2
Data Fusion Architectures
The purpose of fusing multisensor data is to obtain better estimates of geophysical parameters or new information that could not be obtained with any single sensor. Three
basic architectures are utilized in multisensor data fusion processes. These are characterized by the level at which the data are combined or fused - data level, feature level
and decision level (Waltz and Llinas 1990) - and are defined for an identity declaration
process requiring fusion of multisensor data. Although multisensor fusion problems
may not necessarily fall into the identity declaration category, any fusion process is typically characterized by one of these three categories, or a hybrid of these. The architectures are shown in Fig.!.