![](https://csdnimg.cn/release/download_crawler_static/9659601/bg14.jpg)
SENSOR AND DATA FUSION:ATOOL FOR INFORMATION ASSESSMENT AND DECISION MAKING
2
Once the signature-generation processes or observables are known, it is possible
to design a multiple sensor system that captures their unique attributes. Sensors
that respond to signatures generated by different physical phenomena can
subsequently be selected and their outputs combined to provide varying degrees
of immunity to weather, clutter, and diverse countermeasures. Oftentimes, the
data fusion process produces knowledge that is not otherwise obtainable from a
narrow wavelength-band sensor or is more accurate than information gathered
from single sensor systems. An example of the former is the identification of
vegetation on Earth through fusion of hyperspectral visible and infrared data
from space-based sensors such as the Airborne Visible/Infrared Imaging
Spectrometer (AVIRIS). The AVIRIS contains 224 detectors, each with a
spectral bandwidth of approximately 10 nm that cover the 380- to 2500-nm band.
Data fusion also improves the ability of missiles to track and defeat threats. In
this case, accuracy is enhanced by handing off the guidance required for final
missile impact from a lower resolution sensor, optimized for search, to a higher
resolution sensor optimized to find a particular impact area on a target.
The discussion of data fusion that appears in this book is based on the definition
derived from recommendations of the U.S. Department of Defense Joint
Directors of Laboratories Data Fusion Subpanel, namely,
Data fusion is a multilevel, multifaceted process dealing with the
automatic detection, association, correlation, estimation, and
combination of data and information from single and multiple sources to
achieve refined position and identity estimates, and complete and timely
assessments of situations and threats and their significance.
Data fusion consists of a collection of subdisciplines, some of which are more
mature than others. The more mature techniques, such as classical, Bayesian, and
Dempster-Shafer inference; fuzzy logic; pattern recognition using signal
processing algorithms and artificial neural networks; and multisensor, multitarget
tracking, draw on a theoretical apparatus that supports their application. The less
mature techniques are dominated by heuristic and ad hoc methods.
The terms data fusion and sensor fusion are often used interchangeably. Strictly
speaking, data fusion is defined as above. Sensor fusion, then, describes the use
of more than one sensor in a multisensor system to enable more accurate or
additional data to be gathered about events or objects present in the observation
space of the sensors. More than one sensor may be needed to fully monitor the
observation space at all times for a number of reasons: (1) some objects may be
detected by one sensor but not another because of the manner in which signatures
are generated, i.e., each sensor may respond to a different signature-generation
phenomenology; (2) the signature of an object may be masked or
countermeasured with respect to one sensor but not another; (3) one sensor may
be blocked from viewing objects because of the geometric relation of the sensor