TheInternational Workshop onComputer Architecture forMachine Perception andSensing
A DataFusion · 2006
Tenetarchitecture isa two-tier sensor network architecture that provides amodeltoimplement more complex algorithms duetoincorporation ofless resource-restricted nodes. Stargate-class nodescalled masters formtheuppertier while resource-restricted nodes namedmotes compose thelower tier. This paper introduces adata fusion scheme foratenet architecture based onthecorrelation coefficients between dataset extracted fromthemotes. Eachmaster selects four sentinels tocalculate thedirection inwhichanevent hasbeendetected, andthenusesthis dataasabase datatocalculate thecorrelation coefficient forthe incoming data. Theaggregate output isaresult ofa weighted sumofthedatacollectedfrom theN sensors. Theweights arecalculated basedonthecorrelation coefficients. Theaggregated output iscompared with a LinearMeansSquare(LMS)estimator basedon variance. Theproposed schemeachieves good performance. fromsensors spatially distributed through anareaof interest andprovide a better understanding ofthe phenomenon under consideration. Multisensor datafusion refers totheprocess of acquiring, processing andcombining theinformation gathered fromseveral knowledge sources, suchas sensorsand databases, as wellas offers an interpretation ofdetected events [2]. Themainidea behind datafusion isthat complementary information ofthesameeventwillprovide results that aremore accurate. Therefore, thesystemreliability and robustness areimproved. A datafusion scheme must providea system architecture, topology, communication structure, andfusion level. Thefusion levels weredefined byJoint Directors ofLaboratories (JDL) working groupin1986[3]. Thelevels offusion areobject-level, situation-level, orthreat-level. Atthe object level, modelparameters arefusedsuchas velocity, position, andacceleration. Whilethenext level, situation refinement, attempts to infer relationships between objects. Thetoplevel isthreat refinement andtries toinfer anyfuture threats.