Multisensor data fusion is the process of acquiring multiple data sets from multiple sensors with the intent of building a more precise data set. Often considered more accurate than single-sensor data, this type of information fusion has many applications. For example, combining the data from a temperature sensor with a wind chill sensor can help someone inside understand how cold it might feel outside. Aside from meteorological applications, multisensor data analytics can also be applied to environment analysis, transportation management and target tracking.
The many applications of multisensor data fusion show just how useful information fusion can be. When data are coming from multiple sources, specific sets of data can be revised, replaced or cut from the fused data. For instance, a marine biologist interested in tracking whales might use data fusion to monitor factors that he or she thinks might affect whale habits. The end result of multisensor data fusion processes could be a visual map of whale movement related to seawater temperature or other factors. These types of applications rely on many techniques, including physical equipment, algorithms and the related information fusion mathematics.
Sensor technology, mathematical processes and the application of fused data sets all determine the practical application of multisensor data fusion. The technology and processes used to combine integrated data can be thought of as mimicking natural human ability to perceive an environment and make decisions based on the five senses. Technology-based sensors and the related techniques necessary for data fusion might be more specific, however, than human perception.
The combination of these specific sets of data is a defining feature of multisensor data fusion and differentiates information fusion from data integration. Data integration is a large part of the multisensor data fusion process, however, and might be considered a building block for building more advanced data sets. For instance, a sensor may record many different sets of temperatures within a certain period of time and later build a larger set over a longer period of time. This process differs from multisensor data analytics, however, because it does not generally include information from many different sources.
As part of the data fusion process, data integration is inseparable. Without the information provided by strong data integration, there would be no basis for multisensor data fusion. In fact, a common type of multisensor data analytics is low-level data fusion. This process refers to the combination of raw data to create new data sets that are generally expected to be more specific and synthetic than the raw data.
