Multi-sensor data fusion
Multi-sensor data fusion (Engl. Multi-sensor data fusion short often, only data fusion called) refers to the merging and preparation of fragmentary and sometimes contradictory sensor data in a uniform, comprehensible to humans overall picture of the current situation.
The over the data fusion resulting, so-called. Situational then provides the basis for further in-depth decision-making process. The primary area of application of multi-sensor data fusion systems is in the range of command and control systems ( C3I ), but keep the developed there System concepts are increasingly being used in company-wide controlling systems.
Instead of the term multi-sensor data fusion, the English-language specialist literature also uses the abbreviated terms sensor fusion , data fusion and information fusion as an umbrella term that explicitly includes data sources other than sensors. In contrast to the narrower data fusion term that occurs in the context of data warehouse , which deals with the purely information technology merging of incomplete but identically / similarly structured data, the multi-sensor data fusion approach is much more far-reaching in the following aspects:
a) Non-commensurable data sources: A comprehensive overview of the situation often requires the integration of sensors and data sources that are very different not only in terms of their data structure, but also in terms of their content. A number of processing steps are necessary in order to raise the data to a semantic level at which they can actually be combined. For example, radar data must first be processed into flight tracks and combined with information for identification before they can actually be compared with static sources such as a flight plan stored in a database.
b) Information aging: The frequency of the incoming data is usually different, which means that a multi-sensor data fusion system must be able to process information that is of different ages. The age of the data does not only play a role in terms of whether and how relevant it is for the current situation. Rather, old data often have to be extrapolated into the present in order to decide whether the current observations contradict old data or whether a development can be identified. For example, it must be decided whether the objects detected by two radar stations at different positions at a distance of 10 to 15 seconds are the same aircraft, which is only in a different position at the time of the second observation, or whether it is from two different aircraft is to go out.
c) Information weighting : Depending on the design of the sensor system and its local position, information from the respective sensor can be included in the situation report with different weighting. So is z. For example, it can be assumed that, with the appropriate equipment, the on-board sensors of an interceptor deliver a more reliable picture of the close-up situation than radar systems from a further distance. When weighting information, a wide variety of factors, such as system equipment, measuring ranges, scan frequencies, current positions, etc. of distributed sensor systems, must therefore be included. At a higher level, this weighting also occurs when harmonizing data that has already been merged and exchanged with other management systems.
d) Information interpretation: A situation report prepared according to the principle "the whole is more than the sum of the individual parts" requires the partial interpretation of incoming information. On the one hand, consideration must be given to the quality of the sensors used to provide information (some radar systems estimate the speed of an object, others can measure it; high-resolution laser rangefinders will provide more accurate information than radars). On the other hand, it must be taken into account to what extent a shift in the information weighting changes the situation report and what potential danger an incorrect weighting could cause.
see also information fusion